Calling cmake --install . only installs a single configuration.
To install both debug and release artifacts, the install
invocation needs to be done for each configuration.
To keep the Coin instruction code simpler, delegate the looping
over configurations to a custom CMake script, and use it in the
Coin instructions.
Replace all cmake --install calls in the instructions with calls
to either call_host_install.yaml or call_target_install.yaml.
The path to the script depends on whether we are building
qtbase or another module. In the former case the script should
be called from the build dir, otherwise from the install dir.
The other distinction is whether the host or target env prefix
needs to be added.
Task-number: QTBUG-80900
Change-Id: Ied4bf739e2b1a2307f22fc79c1cfad746c8cbc44
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Aka handle CMAKE_INSTALL_PREFIX in a more relocatable way.
The following story inspired this change.
If a user wants to build a Qt repo into a different install prefix
than the usual Qt one, this will fail configuration because we
look for various things like syncqt, qdoc, etc relative to
CMAKE_INSTALL_PREFIX, which will now point to a different location
where none of the above tools are located.
The intent for such a use case is to support building Qt packages with
Conan, which sets a random install prefix when configuring a repo.
The idea is to derive the qt prefix dynamically from the
QtBuildInternals package location. Essentially it's a reverse relative
path from the QtBuildInternalsConfig.cmake file to the install prefix
that was specified when initially configuring qtbase.
Once the dynamic prefix is computed (so we know where the possibly
relocated Qt is), we can find tools like syncqt and qdoc.
This is an initial attempt to support a use case like that.
More design work will probably needed in case if tools / libs need to
be found in a location different than the Qt install prefix (so
support for multiple install prefixes / search paths).
An example of such a case would be when building qtdeclarative and
qtquickcontrols2 as Conan packages in one go. Most likely the
qmltyperegistrar tool will be located in the random install prefix
set by Conan, so building qtquickcontrols2 might fail due to not
finding the tool in the original Qt install prefix.
As to the implementation details, the change does the following:
- Dynamically computes and sets the
QT_BUILD_INTERNALS_RELOCATABLE_INSTALL_PREFIX variable when
find_package()'ing QtBuildInternals. It's an absolute path
pointing to where the relocated Qt is.
- When building qtbase this variable is not yet available (due
to QtBuildInternalsExtra not existing), in that case we set
the variable to the absolute path of CMAKE_INSTALL_PREFIX
(but only for the initial qtbase configuration).
- Remove QT_BUILD_INTERNALS_ORIGINAL_INSTALL_PREFIX which was used
for standalone tests purposes. It's not needed now that we compute
the location of the Qt prefix dynamically.
- The Unixy qt-cmake and qt-cmake-private shell scripts now
use a relative path to find the toolchain file we created.
- The toolchain file also dynamically computes the location of the Qt
packages, and adds them to CMAKE_PREFIX_PATH.
- A lot of existing CMAKE_INSTALL_PREFIX uses are replaced with
QT_BUILD_INTERNALS_RELOCATABLE_INSTALL_PREFIX. This includes finding
tool locations, mkspecs dir, path environment setup for tools, etc.
- Some places still use CMAKE_PREFIX_PATH in the following cases
- When determining paths while configuring qtbase (valid cases)
- When I wasn't sure what the behavior should be, so I left them
as-is (an example is documentation generation, do we want to
install it into the random Conan prefix, or into the main prefix?
Currently it installs in the random prefix).
Note that relocating a Qt installation does not work for non-prefix /
non-installed builds, due to hardcoded paths to include directories
and libraries in generated FooTargets.cmake files.
Task-number: QTBUG-83999
Change-Id: I87d6558729db93121b1715771034b03ce3295923
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Otherwise we will fail to resolve the symbol and will not create
a forwarding header for the class.
Change-Id: I34922d8458bdb994a194108183ac9b9d14530c5e
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Before this patch there were a few ways to build tests
- Configure all tests as part of the repo build
- Configure all tests as part of the repo build, but don't build
tests by default (-DQT_NO_MAKE_TESTS=ON)
- Configure all tests as a standalone project in a separate build
dir using -QT_BUILD_STANDALONE_TESTS=ON
All of the above incur some time overhead due to the necessity
of configuring all tests.
Sometimes you just want to build ONE test (or a few).
To facilitate that use case, a new shell script called
bin/qt-cmake-standalone-test(.bat) can now be used to configure
and build one or more tests.
The script takes one single argument pointing to the desired test
project path and configures a generic template project that sets up
all the necessary Qt CMake private API, afterwards calling
add_subdirectory on the passed in project.
Example
$ path/to/qt/bin/qt-cmake-standalone-test ./tests/auto/gui/image/qicon
or
$ path/to/qt/bin/qt-cmake-standalone-test ./tests/auto/gui/image
After that, simply run 'ninja && ctest' to build and run the test(s).
This is the CMake equivalent of calling qmake on a test .pro file
(or on a tests SUBDIRS .pro file)
There are 3 details worth mentioning.
Due to the add_subdirectory call, the built artifacts will not
be in the top-level build dir, but rather in a nested build_dir.
The script currently can't handle more than one argument
(the path to the project), so you can't pass additional
-DFoo=bar arguments.
If a test uses a 3rd party library (like Threads::Threads)
which was not a public dependency for any of the Qt modules,
configuration will fail saying that the target was not found.
Perhaps we should consider recording these packages when
generating the StandaloneConfig.cmake files.
Change-Id: Icde6ecb839341d34f341d9a19402c91196ed5aa0
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
This allows doing debug_and_release builds with Ninja on all
platforms.
The "Ninja Multi-Config generator" is available starting with CMake
3.17.
Desired configurations can be set via CMAKE_CONFIGURATION_TYPES.
Possible values: "Release, Debug, RelWithDebInfo, MinRelSize".
For example -DCMAKE_CONFIGURATION_TYPES="Release;Debug".
The first configuration is the 'default' configuration which is
built when calling ninja with no arguments.
To build all targets of a certain configuration use "ninja all:Release"
or "ninja all:Debug".
To build all targets in all configurations use "ninja all:all".
Note that the first configuration influences which configuration of
tools will be used when building the libraries for all configurations.
In simple terms, when configured with
-DCMAKE_CONFIGURATION_TYPES="Release;Debug" the release version of moc
is used by AUTOMOC.
When configured with -DCMAKE_CONFIGURATION_TYPES="Debug;Release"
the debug version of moc is used by AUTOMOC.
Framework builds and Ninja Multi-Config don't currently work together
due to multiple bugs in CMake, which ends up generating an invalid ninja
file with duplicate rules. There are also issues with placement of the
debug artifacts.
This will be handled in a follow up patch after CMake is fixed.
Task-number: QTBUG-76899
Change-Id: If224adc0b71b7d1d6606738101536146aa866cd7
Reviewed-by: Qt CMake Build Bot
Reviewed-by: Leander Beernaert <leander.beernaert@qt.io>
Reviewed-by: Cristian Adam <cristian.adam@qt.io>
Rather than tweaking the parser to cover every eventuality with corner
case lines that could cause incorrect header files to be created then
the means to suspend/resume the processing of a file is added.
This enables us to have it skip over the template line that is causing
a QList header to be created as part of the QtGui headers. This patch
includes the fix to solve this in addition.
Fixes: QTBUG-68129
Change-Id: I751646c4b20a4434347c149ae5e6dcb6e7618853
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
This broke when QLatin1Literal got deprecated in change 45373c1924
Both hunks in this patch are needed. The hunk in syncqt.pl removes the
QT_DEPRECATED_X(...) macro if it appears solely on a line in source code,
the second hunk inserts a newline after the QT_DEPRECATED_X(...) macro usage
to trigger that code path in the Perl script.
Before/after comparison of the headers generated in include/QtCore:
```
% diff ~/old.txt ~/new.txt
105a106
> QLatin1Literal
```
Change-Id: I468dd2dd54bf115521ed82c6182236905556f568
Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
In Windows-msys syncqt.pl expects CRLF line endings, and does not
work correctly with LF. syncqt.pl was fixed to be line-ending-agnostic.
Task-number: QTBUG-77192
Change-Id: Ie8029238bdd580bcf042ede0d0f64d5f01488406
Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
Modified a regular expression in syncqt.pl so that the special case of
a class beginning with another class does not lead to the exclusion of
the first one. This affects the generation of the install target for
generated class headers of Qt modules. Now the expression verifies the
class names are not identical.
Fixes: QTBUG-71323
Change-Id: I210b4d4c3ed64cf189594b95b10aa0e8495a19d2
Reviewed-by: Jörg Bornemann <joerg.bornemann@qt.io>
Previously the cmake install prefix was added twice during a non-prefix
build, which evaluated to an incorrect path.
Set another absolute dir variable, and use that instead.
Change-Id: I73099510dadc8f401d5a763f21840c9671686c10
Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
Reviewed-by: Qt CMake Build Bot
Similar to the shell script, this makes it easier to invoke cmake by
just extending your PATH.
Change-Id: I1f83e93c507032f6ecf4838c11d62f49f4d93ed6
Reviewed-by: Qt CMake Build Bot
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
This gets us a step into the direction of convenience that qmake
offered:
* QtBase is configured with a long command line (especially when
cross-compiling)
* Afterwards application developers (or other module builds) can
just use qmake && make
By generating a toolchain file we can capture vcpkg and toolchain
chain-loading and a shell script can take care of providing the prefix
path.
Change-Id: Ided81f5432cab862306f2bea86cfe8e56adf71b0
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
Injected headers were made relative to MODULE_BASE_OUTDIR by syncqt
and made absolute by resolving against REAL_MODULE_BASE_OUTDIR.
This breaks for modules that reside outside the original Qt source
tree (if the directory depth doesn't coincidentally match).
Now, we resolve injected headers against build_basedir, which is
REAL_MODULE_BASE_OUTDIR. To emphasize the equivalence of
REAL_MODULE_BASE_OUTDIR and syncqt's build_basedir, use the former for
syncqt's -output argument.
This commit amends 2aa779e8.
Fixes: QTBUG-70587
Change-Id: I2935d87d7ee681fa4aa795a270b94ab7a43abe59
Reviewed-by: Dominik Holland <dominik.holland@pelagicore.com>
Reviewed-by: Oliver Wolff <oliver.wolff@qt.io>
It is possible that "syncqt -minimal" is run after a module's initial run
(for example if a host tool relies on a module, but cannot use it
directly and thus adds minimal_syncqt to its .pro file). In this case, the
old directory content should not be wiped.
Fixes: QTBUG-59319
Change-Id: I83767eff0ef74bcefae5efa9b18b7ab3724138e5
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
in non-prefix builds, the forwarding headers always end up in qtbase's
build dir, while the injected headers always live in the build dir of
the module they belong to. to deal with that, we now record the target
path relative to the module root dir instead of relative to the base
directory of the forwarding header itself.
Fixes: QTBUG-70056
Change-Id: Ic4346148a125b13e2610f6965cdf4f5266ac763e
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
following the same mechanism as qpa headers, one can specify a list of
regexes in the @private_headers variable in sync.profile.
Change-Id: I5de0284e639ad283561f54dba7dda8c6437b23f8
Reviewed-by: Johan Helsing <johan.helsing@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
this variable hasn't been used in any syncqt.profile for many years, and
if it would, it would cause quite some side effects.
Change-Id: Ia936eff8f6a2a801fc644eee991821165d51dc1f
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
don't put them into GENERATED_HEADER_FILES, as they obviously cannot be
found in a pre-synced source dir. instead, let the injection code itself
add them to INJECTED_HEADER_FILES.
Task-number: QTBUG-67813
Change-Id: Id2a7c565b14fcba8aba9d1dd8b1dd39c586d0d91
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
these are actually redundant with INJECTIONS.
Change-Id: I0a71930401e00d30c9898b4d958de5e89c496d18
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
instead of relying on more or less accidental qmake behaviors regarding
the base dir for relative paths (esp. if a file does not exist yet),
make everything explicit. to that effect, clearly define the base tree
(source or build) for every syncqt-generated variable, and write only
in-tree relative paths to the variables. on the receiving end, resolve
the paths as soon as headers.pri was read.
Task-number: QTBUG-67111
Change-Id: I32ae5760fb62ebc650fdb69e46aac786a8141564
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Alexandru Croitor <alexandru.croitor@qt.io>
creating private deprecated forwarding headers is rather pointless, as
keeping source compatibility for private headers is unnecessary.
consequently, the feature wasn't used anywhere. remove it to reduce
complexity.
Change-Id: I6202e5141f0387615f90efa0ec707ac208419227
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
it's been a no-op (always on anyway) for quite some time now.
Change-Id: I9611ca73ddaef4422a16d85338e1dae473a50fbf
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
that parameter is never set, and if it was, it wouldn't work: the
recursively found files were not qualified with the subdirectory.
Change-Id: I24a3f74de3545c19bcd175c24ad1daf2af77a720
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
it needs to fall back to -outdir if only the latter is specified, as
otherwise funny things will happen.
this didn't affect the calls in qt_module_headers.prf (both options
used) and the unix configure script (neither option used), but it
affected configure.bat, and possible callers outside qt's build system,
which apparently includes our packaging scripts.
amends ede6c44756.
Task-number: QTBUG-66626
Change-Id: I08633033c04152616545501490c9b3fb939cab6c
Reviewed-by: Lionel CHAZALLON <longchair@hotmail.com>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
in non-prefix configs, one has to differentiate between the module's own
build dir and qtbase's build dir, because the forwarding headers are
placed in -outdir under include/, while the actual headers end up in the
real build dir under src/.
Change-Id: I1d8ac904556b354bd113995316ba11dd6560a70d
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
don't pass arguments, as the sourced sync.profile's make assumptions
about the global variables' presence anyway.
don't return a value, as it's ignored anyway.
fix the documentation (also that of locateSyncProfile()).
Change-Id: I7527546718b12737d7a4551a834d7f0ca26a8a7e
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
For source packages that don't have a .git subdirectory, syncqt is
executed before configure, with outdir set to srcdir, and this
caused path misalignments for injected headers in qt_module.prf
when generating makefile rules.
The fix is to change syncqt to always output injected header
paths relative to the source dir.
Task-number: QTBUG-64539
Change-Id: Ia2296e44494093dbf124729062f430ad6fca7262
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
Bundle data source files which don't exist at qmake time need to be
handled specially. This also required splitting the generated list of
public headers, as was already done for private ones.
Task-number: QTBUG-60413
Change-Id: I97acfa88622da6b73839b8f976f73ace3cb10223
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@qt.io>
the target of the injected forwarding header doesn't exist at qmake
time, as it is generated by an extra compiler, so the touch() calls in
qt_module_headers.prf would fail.
the error scenario described in ce942a226 is not applicable to
gui/vulkan, as no bootstrapped modules are involved. therefore, we can
just suppress the timestamping.
Change-Id: I1c9b6fcdf3717069fdbb654e3cb5d73b199192f4
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>
Reviewed-by: Joerg Bornemann <joerg.bornemann@qt.io>
Reviewed-by: Laszlo Agocs <laszlo.agocs@qt.io>
it makes no sense to version them, as they contain only #defines anyway.
it also removes the need to special-case their location in shadow builds
with pre-synced headers, which we actually failed to do anyway.
Task-number: QTBUG-56286
Change-Id: I4ea717f7be56494cfea0572389bea173d7470b6e
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
it makes no sense to check them, as they contain only #defines anyway.
Change-Id: I8b36139ee19471de0654c5eb3af262d0389a72f7
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
To match correctly (only) .h files, a regex needs to end in \.h$
Some of them missed the \, one missed the $.
(The last also had a legitimate .* before its misunescaped .)
One pair matched _p.h and _pch.h, which could be combined.
Change-Id: I7539a28eb7017cd0f1b36c72e05108e03a68a952
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
The check for Q_... macros used a regex that didn't allow digits; it
thus failed to match Q_DECLARE_SHARED_NOT_MOVABLE_UNTIL_QT6().
Change-Id: I3f9339349aa21e2fea04a7f53d9d8e0903e4c65b
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
This macro expands into a static_assert and can be used to
trigger a compile error if a certain feature is not available
when trying to compile some code.
This is especially useful to protect against accidental inclusion
of headers that implement functionality related to a feature.
Change-Id: I456c55b989ce5f35f3af0e13c1886a85c23dfe29
Reviewed-by: Edward Welbourne <edward.welbourne@qt.io>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
This change implements the required infrastructure to modularize the new
configuration system.
This requires a hierarchy of configuration files, both for handling
multiple repositories and for individual modules inside the same
repository.
When configuring, they all need to get loaded first, as command line
processing needs to know about all possible command line options.
When the command line has been processed, the individual configuration
files need to get processed one after the other and independently from
each other.
Configure is now automatically invoked when building the a project
tree's "root" project; this works with both modular and top-level builds
of Qt (the latter with an according change in the super repo). As an
immediate consequence, the -skip option moves to the super repo with a
different implementation, as configuration is now done after the repo
list is determined. The option belongs there anyway.
This commit also adds an optional testDir entry to the json file. Like
this, we can still have all configure tests in qtbase/config.tests and
the configuration file in, e.g., corelib can reference those.
The files section can now be left out as long as a 'module' entry is
present, specifying the module name. The names of the files to generate
can then be deduced from that name. We still need to be able to specify
names directly for the global configuration files.
qtConfig() now also queries features which are module-specific. As it is
sometimes necessary to query the configuration of modules which should
not be actually linked (and cannot in the case of subdirs projects), the
new variable QT_FOR_CONFIG which allows specifying configuration-only
dependencies is introduced.
Done-with: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Change-Id: Id1b518a3aa34044748b87fb8fac14d79653f6b18
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
don't bother opening headers.pri files, as they won't match anyway.
Change-Id: I4d923266dabf1c9684fba4086f55bc24d76d23c5
Reviewed-by: Oswald Buddenhagen <oswald.buddenhagen@theqtcompany.com>
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
the limitation to public headers seemed pretty arbitrary.
Change-Id: I0f2290fe09c8a6a70d056ec250b902be7049790e
Reviewed-by: Lars Knoll <lars.knoll@qt.io>
