This reverts commit f52bcaa27f.
It did not pass CI, and was merged anyway because there were two CI
errors in the same cygwin job. The other error was not the fault of this
commit, and since cygwin errors were glossed over because they were
"expected", the presence of a new error *added* by this commit was
overlooked.
Per the meson development policy, PRs which result in CI errors
can/should be reverted at will, no questions asked.
In commit f52bcaa27f a few issues were
added:
- doc typo
- imports for utils.universal are not intended to be directly used, it's
an internal wrapper that exists solely to make mesonlib work well as
it always did while simultaneously allowing `meson --internal`
codepaths to avoid importing anything other than an extremely stripped
down core
- type annotation specific import was imported at runtime scope
This commit adds a new keyword arg to extension_module() that enables
a user to target the Python Limited API, declaring the version of the
limited API that they wish to target.
Two new unittests have been added to test this functionality.
Performed using https://github.com/ilevkivskyi/com2ann
This has no actual effect on the codebase as type checkers (still)
support both and negligible effect on runtime performance since
__future__ annotations ameliorates that. Technically, the bytecode would
be bigger for non function-local annotations, of which we have many
either way.
So if it doesn't really matter, why do a large-scale refactor? Simple:
because people keep wanting to, but it's getting nickle-and-dimed. If
we're going to do this we might as well do it consistently in one shot,
using tooling that guarantees repeatability and correctness.
Repeat with:
```
com2ann mesonbuild/
```
ExternalProgram and CustomTarget have some use cases for producing
subclassed interpreter holders with more specific types and methods. In
order for those subclasses to properly refer to their held_object, we
need a shared base class that is still generic, though bound.
For the derived held objects, inherit from the base class and specify
the final types as the module-specific type.
This fixes two issues in constructing the default installation path
when install_dir is not specified:
- inside a subproject, install_data() would construct the destination
path using the parent project name instead than the current project
name,
- when specifying preserve_path, install_data() would construct the
destination path omitting the project name.
Fixes#11910.
When the pkgconfig module is imported, but not used, it will insert None
on the end of the devenv list. This list is not expected to contain
None, and causes Meson to crash. This can happen in complex build setups
(reported from mesa), where pkgconfig is only used in some
configurations
Fixes: #12032
We already import a bunch of objects directly from ..build but don't use
them nearly as much as we can. This resulted both in longer lines and s
minor performance difference since python has to resolve the name
binding the long way. There's no reason not to rewrite these names to
use the direct imports.
Found while investigating the fact that Executable was imported but
never used. It's easier to just use it.
This detects cases where module A imports a function from B, and C
imports that same function from A instead of B. It's not part of the API
contract of A, and causes innocent refactoring to break things.
On windows, meson would mix posix and windows dir separators in the
computed PACKAGE_RELATIVE_PATH.
Here we force posix directory separator even on Windows. This matches
the CMake behavior and fixes interpretation of the resulting path.
Fixes#6955Fixes#9702
By default, clang-cl based environments use rc.exe as resource
compiler. However, when cross compiling with clang-cl, one might
want to use llvm-rc instead.
Try to detect llvm-rc based on the output from "$CMD /?".
This requires a very recent llvm-rc; previosly, the output of
"/?" with llvm-rc was very generic and didn't explicitly indicate
that it actually was llvm-rc. This was changed in
bab6902eba
which will be included in the upcoming LLVM 17.0.0 release.
Contrary to the other regexes, don't include the preceding parts
of the line in the log printout, as it includes an unhelpful
"OVERVIEW:" prefix.
It's currently impossible to inject extra clang arguments when using
bindgen, which is problematic when cross compiling since you may need
critical arguments like `--target=...`. Because such arguments must be
passed after the `--` it's impossible to inject them currently without
going to something like a wrapper script.
Fixes: #11805
This reverts commit a2def550c5.
This results in a 2k line file being unconditionally imported at
startup, and transitively loading two more (for a total cost of 2759
lines of code), and it's not clear it was ever needed to begin with...
This lessens the amount of code imported at Meson startup by mapping
each dependency to a dictionary entry and using a programmable import to
dynamically return it.
Minus 16 files and 6399 lines of code imported at startup.
We can check something's subtype using properties, without importing the
module up front and doing isinstance checks on specific subclasses of
the interface -- or worse, solving cyclic imports by doing the import
inside the function. ;)
msetup.py doesn't need to know the gory details of PkgConfigDependency,
or directly import it at program startup. It's also slightly wasteful to
generate a devenv for the -uninstalled directory when a project doesn't
even, in the end, use the pkgconfig module anyway.
This is a pretty common pattern in python (the standard library uses it
a ton): A class is created, with a single private instance in the
module, and then it's methods are exposed as public API. This removes
the need for the global statement, and is generally a little easier to
reason about thanks to encapsulation.
The concept of merge_file intrinsically requires some GNU-specific
functionality, so let's emit a useful error message during
configuration, when we don't have that.
The relevant GNU gettext versions date back to around 2015 so *probably*
anyone has that too, but we may as well verify that while we are here.
There are a number of implementations for msgfmt, supporting various
options. The simplest, and most common, use case is to compile .po files
into .mo files, and this should be able to work on gettext
implementations other than the GNU one.
The problem is that we were passing some pretty portable arguments in an
unportable manner. The `-o` option-argument and its associated argument
came after the input file operand, which violates the POSIX Utility
Syntax Guidelines, and happens to not be supported by Solaris gettext.
The GNU gettext doesn't care; GNU invented GNU argument permutation.
Switch the order around so that our use respects the POSIX style.
In #11761 it turned out that we failed to correctly handle all
compiler.sizeof API changes in an old commit, breaking use of the
module. And mypy could have caught this for us, except that the module
is neither typed nor checked in CI. Partially solve this by adding lots
of type annotations, greatly reducing the number of mypy errors in this
file from 35 down to 12.
This doesn't accept a dict, only an actual ConfigurationData object. Due
to the way we poke at it, a dict can sort of work anyway, but might not
if the internal layout isn't exactly correct. This is evidenced by the
way we make the dict values be hard-to-read tuples containing emptiness,
because that's how ConfigurationData objects handle descriptions.
Simplify and make the seed dictionary readable, then actually convert it
into a real ConfigurationData. Bonus: this now passes type checking.
It's shorter and more descriptive. Although we always enforce the same
rules either way, a unified decorator is one less line of code for each
location, and also tells you how many "too few" arguments you *did*
pass.
Xavier Claessens
Tristan Partin
Charles Brunet
Eli Schwartz
Andrew McNulty
Dylan Baker
Daniele Nicolodi
Matthieu Rogez
Martin Storsjö
L. E. Segovia
Volker Weißmann