New features

Meson 1.2.0 was released on 17 July 2023

Added Metrowerks C/C++ toolchains

Added support for the Metrowerks Embedded ARM and Metrowerks Embedded PowerPC toolchains (https://www.nxp.com/docs/en/reference-manual/CWMCUKINCMPREF.pdf).

The implementation is somewhat experimental. It has been tested on a few projects and works fairly well, but may have issues.

Added str.splitlines method

str.splitlines() can now be used to split a string into an array of lines.

generator.process(generator.process(...))

Added support for code like this:

gen1 = generator(...)
gen2 = generator(...)
gen2.process(gen1.process('input.txt'))

Extra files keyword in declare_dependency

declare_dependency have a new extra_files keyword, to add extra files to a target. It is used mostly for IDE integration.

Added a new '--genvslite' option for use with 'meson setup ...'

To facilitate a more usual visual studio work-flow of supporting and switching between multiple build configurations (buildtypes) within the same solution, among other reasons, use of this new option has the effect of setting up multiple ninja back-end-configured build directories, named with their respective buildtype suffix. E.g. 'somebuilddir_debug', 'somebuilddir_release', etc. as well as a '_vs'-suffixed directory that contains the generated multi-buildtype solution. Building/cleaning/rebuilding in the solution now launches the meson build (compile) of the corresponding buildtype-suffixed build directory, instead of using Visual Studio's native engine.

gnome.generate_gir() now supports env kwarg

gnome.generate_gir() now accepts the env kwarg which lets you set environment variables.

More data in introspection files

  • Used compilers are listed in intro-compilers.json
  • Informations about host, build and target machines are lister in intro-machines.json
  • intro-dependencies.json now includes internal dependencies, and relations between dependencies.
  • intro-targets.json now includes dependencies, vs_module_defs, win_subsystem, and linker parameters.

Machine objects get kernel and subsystem properties

Meson has traditionally provided a system property to detect the system being run on. However this is not enough to reliably differentiate between e.g. an iOS platform from a watchOS one. Two new properties, namely kernel and subsystem have been added so these setups can be reliably detected.

These new properties are not necessary in cross files for now, but if they are not defined and a build file tries to access them, Meson will exit with a hard error. It is expected that at some point in the future defining the new properties will become mandatory.

default_options and override_options may now be dictionaries

Instead of passing them as default_options : ['key=value'], they can now be passed as default_options : {'key': 'value'}, and the same for override_options.

New override of find_program('meson')

In some cases, it has been useful for build scripts to access the Meson command used to invoke the build script. This has led to various ad-hoc solutions that can be very brittle and project-specific.

meson_prog = find_program('meson')

This call will supply the build script with an external program pointing at the invoked Meson.

Because Meson also uses find_program for program lookups internally, this override will also be handled in cases similar to the following:

custom_target(
  # ...
  command: [
    'meson',
  ],
  # ...
)

run_command(
  'meson',
  # ...
)

run_target(
  'tgt',
  command: [
    'meson',
    # ...
  ]
)

Find more specific python version on Windows

You can now use python3.x, where x is the minor version, to find a more specific version of python on Windows, when using the python module. On other platforms, it was already working as python3.x is the executable name.

Python module can now compile bytecode

A new builtin option is available: -Dpython.bytecompile=2. It can be used to compile bytecode for all pure python files installed via the python module.

rust.bindgen allows passing extra arguments to rustc

This may be necessary to pass extra cfgs or to change warning levels.

Support for defining crate names of Rust dependencies in Rust targets

Rust supports defining a different crate name for a dependency than what the actual crate name during compilation of that dependency was.

This allows using multiple versions of the same crate at once, or simply using a shorter name of the crate for convenience.

a_dep = dependency('some-very-long-name')

my_executable = executable('my-executable', 'src/main.rs',
  rust_dependency_map : {
    'some_very_long_name' : 'a',
  },
  dependencies : [a_dep],
)

A machine file may be used to pass extra arguments to clang in a bindgen call

Because of the way that bindgen proxies arguments to clang the only choice to add extra arguments currently is to wrap bindgen in a script, since the arguments must come after a --. This is inelegant, and not very portable. Now a bindgen_clang_arguments field may be placed in the machine file for the host machine, and these arguments will be added to every bindgen call for clang. This is intended to be useful for things like injecting --target arguments.

This can already be be worked around by creating declare_dependency() objects to pass to the dependencies keyword, but this cuts out the middle man.

Rust now supports the b_ndebug option

Which controls the debug_assertions cfg, which in turn controls debug_assert!() macro. This macro is roughly equivalent to C's assert(), as it can be toggled with command line options, unlike Rust's assert!(), which cannot be turned off, and is not designed to be.

Wildcards in list of tests to run

The meson test command now accepts wildcards in the list of test names. For example meson test basic* will run all tests whose name begins with "basic".

meson will report an error if the given test name does not match any existing test. meson will log a warning if two redundant test names are given (for example if you give both "proj:basic" and "proj:").

New for the generation of Visual Studio vcxproj projects

When vcxproj is generated, another file vcxproj.filters is generated in parallel. It enables to set a hierarchy of the files inside the solution following their place on filesystem.

The results of the search are