* Per-config values were added to `AUTO*_EXECUTABLE`.
* Dependency order was refactored for `cmake_autogen` and `cmake_autorcc` to avoid unnecessary rebuilds.
* A new parameter was added for `cmake_autogen` and `cmake_autorcc` to specify the config name of the `auto*_executable` to be used.
* Add `AUTOGEN_BETTER_GRAPH_MULTI_CONFIG` target property to change the behavior of the dependency graph.
* The timestamp target is split into three targets for per-config to avoid redundant `mocs_compilation` builds when `AUTOGEN_BETTER_GRAPH_MULTI_CONFIG` is ON
* Per-config `DEP_FILE_RULE_NAME` values were added to `AutogenInfo.json` for `Multi-Config` usage.
* Some functions were refactored to avoid code duplication.
This commit reimplements fddd0f0443b4ce81d61f15ee1b2f13105967b25a
Fixes: #20074
Changing the `timestamp` file to `timestamp_$<CONFIG>` causes some user
projects to break when using Qt versions older than 6.6.
Revert commit fddd0f0443 (Autogen: AUTO*_EXECUTABLE: add support for
per-config values, 2023-06-14, v3.28.0-rc1~96^2~1) pending further
investigation.
Issue: #20074
* Per-config values were added to `AUTO*_EXECUTABLE`.
* Dependency order was refactored for `cmake_autogen` and `cmake_autorcc` to
avoid unnecessary rebuilds.
* A new parameter was added for `cmake_autogen` and `cmake_autorcc` to specify the config name of the `auto*_executable` to be used.
* The timestamp target was split into three targets for per-config to avoid redundant `mocs_compilation` builds.
* Per-config `DEP_FILE_RULE_NAME` values were added to `AutogenInfo.json` for `CMAKE_CROSS_CONFIG` usage.
* Some functions were refactored to avoid code duplication.
Fixes: #20074
When merging a new set of uic options with a base set of options,
such as when a source file has a non-empty AUTOUIC_OPTIONS
property, the test for whether to support options starting with
two hyphens was only testing if the Qt major version was exactly
5 rather than at least 5. That had the effect of preventing such
options from being merged correctly when using Qt6, so new and
base options would both be present instead of the single merged
option from the source file properties. Update the version check
and rename function arguments to avoid misrepresenting what
they mean.
Note that uic accepts long option names with a single hyphen too.
See Qt docs for QCommandLineParser::ParseAsLongOptions
for confirmation.
Revert commit e5ec0e52f4 (AUTOUIC: Fix generating of dependency rules
for UI header files, 2021-07-22, v3.21.1~8^2) because it caused
regressions. For example, changing one C++ source can now cause many
others to rebuild. Revert the change pending further investigation.
Fixes: #22531
Issue: #16776
We could not rely on .ui files when generating the ninja rules
for the generated UI header files. .ui files might be added to the
target sources but never processed by AUTOUIC afterward, since UI
header files are never included in a source code. Instead of adding
dependency rules based on the .ui files, this approach scans
non-generated source files for includes of the UI header files,
as AUTOUIC does. This gives the consistent set of UI header files
at configure time, that could be used to generate byproducts rules
for the AUTOUIC. Also, the path to the generated UI header file depends
not on the .ui file location but on the include line is used in source
files.
Fixes: #16776
#pragma once is a widely supported compiler pragma, even though it is
not part of the C++ standard. Many of the issues keeping #pragma once
from being standardized (distributed filesystems, build farms, hard
links, etc.) do not apply to CMake - it is easy to build CMake on a
single machine. CMake also does not install any header files which can
be consumed by other projects (though cmCPluginAPI.h has been
deliberately omitted from this conversion in case anyone is still using
it.) Finally, #pragma once has been required to build CMake since at
least August 2017 (7f29bbe6 enabled server mode unconditionally, which
had been using #pragma once since September 2016 (b13d3e0d)). The fact
that we now require C++11 filters out old compilers, and it is unlikely
that there is a compiler which supports C++11 but does not support
#pragma once.
We used to store information for the _autogen target in a CMake script
file AutogenInfo.cmake, which was imported by a temporary cmake instance in
the _autogen target. This introduced the overhead of creating a temporary
cmake instance and inherited the limitations of the CMake language which
only supports lists.
This patch introduces JSON files to pass information to AUTORCC and
autogen_ targets. JSON files are more flexible for passing data, e.g. they
support nested lists.
The patch has the side effects that
- AutogenInfo.cmake is renamed to AutogenInfo.json
- AutogenOldSettings.txt is renamed to AutogenUsed.txt
- RCC<qrcBaseName><checksum>Info.cmake is renamed to
AutoRcc_<qrcBaseName>_<checksum>_Info.json
- RCC<qrcBaseName><checksum>.lock is renamed to
AutoRcc_<qrcBaseName>_<checksum>_Lock.lock
- RCC<qrcBaseName><checksum>Settings.txt is renamed to
AutoRcc_<qrcBaseName>_<checksum>_Used.txt
Provide a standardized way to handle the C++ "standard" headers
customized to be used with current CMake C++ standard constraints.
Offer under directory `cm` headers which can be used as direct
replacements of the standard ones. For example:
#include <cm/string_view>
can be used safely for CMake development in place of the `<string_view>`
standard header.
Fixes: #19491
By splitting `JobProbeDepsT` into two independent `JobProbeDepsMocT`,
`JobProbeDepsUicT`, the moc and uic job generation will happen concurrently.
This should improve the execution speed, when both AUTOMOC and AUTOUIC are
enabled.
Parent directory names for moc and uic output files are are collected in
a `std::unordered_set<std::string>` and get created in a batch.
Beforehand we tried to create the parent directory for each output file.
This led to duplications of `MakeDirectory` calls when there were multiple
output files in the same directory (which is the case usually).
An old workaround for `std::allocator_traits<>::value_type` lints from
IWYU on `std::vector<>` usage breaks IWYU's handling of `<memory>`.
Convert the workaround to use the same approach we already use for a
workaround of `std::__decay_and_strip<>::::__type` lints. Then update
the `<memory>` inclusions to follow the now-correct IWYU lints.
To speed up the `AUTOGEN` configuration process, evaluate the compiler
features only once. The feature evaluation result is stored in the new class
`cmQtAutoGen::CompilerFeatures`, and the instance is shared by using
`std::shared_ptr`.
The `.qrc` parsing functions in `cmQtAutoGen` are no longer needed outside
the private implementation of `cmQtAutoGen::RccLister`. This makes them
private, too.
The new `cmQtAutoGen::RccLister` class provides a simple interface to
list files in a `.qrc` resources file by either calling `rcc` with
list options or by parsing the resources file content.
For multi configuration generators AUTOMOC generates the
moc files that are included in `mocs_compilation.cpp`
in `AUTOGEN_BUILD_DIR/include_$<CONFIG>/`. By doing so each
configuration reads different moc files when compiling
`mocs_compilation.cpp`. Since we do not (need to) rewrite
`mocs_compilation.cpp` on a configuration change anymore,
the files also does not need to be recompiled anymore.
Not having to rewrite and recompile `mocs_compilation.cpp`
on a configuration change anymore was the main objective of
this patch.
In a similar fashion AUTORCC generates a `qrc_BASE_CMAKE.cpp`
file below `AUTOGEN_BUILD_DIR/include_$<CONFIG>/` and
`qrc_BASE.cpp` becomes a mere wrapper that includes this
actuall rcc output file (when using multi configuration
generators).
The template files `Modules/AutoRccInfo.cmake.in` and
`Modules/AutogenInfo.cmake.in` were removed in favor
of writing the info `.cmake` files manually.
Closes#17230
This introduces concurrent thread processing in the `_autogen`
target wich processes AUTOMOC and AUTOUIC.
Source file parsing is distributed among the threads by
using a job queue from which the threads pull new parse jobs.
Each thread might start an independent ``moc`` or ``uic`` process.
Altogether this roughly speeds up the AUTOMOC and AUTOUIC build
process by the number of physical CPUs on the host system.
The exact number of threads to start in the `_autogen` target
is controlled by the new AUTOGEN_PARALLEL target property which
is initialized by the new CMAKE_AUTOGEN_PARALLEL variable.
If AUTOGEN_PARALLEL is empty or unset (which is the default)
the thread count is set to the number of physical CPUs on
the host system.
The AUTOMOC/AUTOUIC generator and the AUTORCC generator are
refactored to use a libuv loop internally.
Closes#17422.
We used to detect the `rcc` features before every `rcc` list invocation
wich resulted in `rcc` be called twice for every listing operation.
Now we detect the `rcc` list capabilities once during configuration and
pass it to the cmake_autorcc target in the info file.
This is a large commit that serves multiple purposes
- Iterate source files only once and store all extracted
information in a cmQtAutogenDigest class that can be reused.
This is brings speed improvements because several properties
are only evaluated once. More that that it helps to avoid
duplication of code with non trivial files property checks.
- Fix the Visual Studio generator to use PRE_BUILD when possible.
- Convert `for( ... )` loops to C++11 range base loops where possible
(cmQtAutogen*.cxx only).
- String concatenation optimizations.
