In most cases instead pass `for_machine`, the name of the relevant
machines (what compilers target, what targets run on, etc). This allows
us to use the cross code path in the native case, deduplicating the
code.
As one can see, environment got bigger as more information is kept
structured there, while ninjabackend got a smaller. Overall a few amount
of lines were added, but the hope is what's added is a lot simpler than
what's removed.
determine_rpath_dirs() can return paths to external dependencies not
in the build dir and passing them first as a link path leads to
g-ir-scanner for example linking against the already installed library
instead of the just built one.
This was reported in g-i: https://gitlab.gnome.org/GNOME/gobject-introspection/issues/272
and comes up quite often when a library adds some new symbols which aren't present in the
system library, which then makes linking fail.
The first place where the order is changed is _scan_gir_targets(), which looks like an unintentional
change in the refactoring in 8377ea45aa
The second place in _get_link_args() has always been that way and only the rpath order is changed,
but it looks to me as if the same rules should apply here too.
Instead use coredata.compiler_options.<machine>. This brings the cross
and native code paths closer together, since both now use that.
Command line options are interpreted just as before, for backwards
compatibility. This does introduce some funny conditionals. In the
future, I'd like to change the interpretation of command line options so
- The logic is cross-agnostic, i.e. there are no conditions affected by
`is_cross_build()`.
- Compiler args for both the build and host machines can always be
controlled by the command line.
- Compiler args for both machines can always be controlled separately.
Although `gtkdoc` function has support for `c_args` argument[0], it
produces warning messages due to missing string in the permitted
arguments list.
[0] https://github.com/mesonbuild/meson/pull/4192
It is possible for compiler flags to include special characters, such as
double quotes which are needed to define macros with -D options. Since
gtkdoc-scangobj uses shlex.split to split arguments passed to --cc,
--ld, --cflags, --ldflags into lists, we can safely use shlex.quote to
properly quote arguments for these options.
When passing static libraries to gtkdoc, they are also appended as
shared libraries to ldflags, which makes the process to fail.
This has been changed to only append shared libraries to ldflags.
Fixes#3935
gtk-doc for autotools has the concept of module version, that is used to define
the module install path and the devhelp2 basename.
Add a `module_version` parameter to gnome.gtkdoc to replicate the same behavior.
Updated the test checking that the install_dir is properly computed (if not
passed), and that the .devhelp2 file has proper name.
https://gitlab.gnome.org/GNOME/gtk-doc/blob/GTK_DOC_1_29/buildsystems/autotools/gtk-doc.make#L269
gtkdoc-scangobj also accepts compiler arguments. In the same way
that include_directories includes directories, the new c_args
parameter also appends compiler arguments.
One of the gtkdoc's steps calls to gtkdoc-scangobj that also accepts
compiler arguments by using the cflags option.
Compiler arguments from dependencies are also appended now.
The regression was introduced in my recent refactoring of
that method (8377ea4).
This commit simply restores the ordering of the generated
scan_command, ensuring `-lasan` and other internal linker
flags come before `--library` or `--program`
g-ir-scanner is very picky about the flags that it can accept, so the
build fails on macOS if you have Framework external dependencies,
which add -F and -framework arguments.
Also fix incorrect de-duping of -framework arguments for gtkdoc.
* Fix flake8 whitespace reports
$ flake8 | grep -E '(E203|E221|E226|E303|W291|W293)'
./mesonbuild/coredata.py:337:5: E303 too many blank lines (2)
* Fix flake8 'variable assigned value but unused' reports
$ flake8 | grep -E F841
./mesonbuild/modules/gnome.py:922:9: F841 local variable 'target_name' is assigned to but never used
* Fix flake8 'imported but unused' reports
$ flake8 | grep F401
./mesonbuild/compilers/__init__.py:128:1: F401 '.c.ArmclangCCompiler' imported but unused
./mesonbuild/compilers/__init__.py:138:1: F401 '.cpp.ArmclangCPPCompiler' imported but unused
./mesonbuild/modules/__init__.py:4:1: F401 '..mlog' imported but unused
PR #3717 imports ARMCLANG compilers in __init__, but does not add them to
__all__, so they are not re-exported by the compilers package like
everything else.
* More details about flake8 in Contributing.md
Mention that Sider runs flake8
Suggest seting flake8 as a pre-commit hook
Those tools use our arguments to build a file and execute it to
introspect it at runtime. However, they do not know that you can pass
the full path to the library to use, and ignore the arguments.
The long-term fix for this is to have them output a .c file that Meson
will build for them, which they can then run, but that will require
upstream changes:
https://gitlab.gnome.org/GNOME/gtk-doc/merge_requests/1
Closes https://github.com/mesonbuild/meson/issues/3774
When an older version of the library being built is installed in the
same prefix as external dependencies, we have to be careful to construct
the linker or compiler command line. If a -L flag from external
dependencoes comes before a -L flag pointing to builddir, it is possible
for the linker to load older libraries from the installation prefix
instead of the newly built ones, which is likely to cause undefined
reference error.
Since the order of dependencies is not significant, we cannot expect
internal dependencies to appear before external dependencies when
recursively iterating the list of dependencies. To make it harder to
make mistakes, linker flags come from internal and external
dependencies are now stored in different order sets. Code using
_get_dependencies_flags are expected to follow the order when
constructing linker command line:
1. Internal linker flags
2. LDFLAGS set by users
3. External linker flags
It is similar to what automake and libtool do for autotools projects.
In GLib when running "ninja gio-doc" without doing a full build first,
it won't build libraries and the generated gio-scan.c gets linked
against system libgio instead.
This fix 2 bugs: gtkdoc() was not passing 'depends' variable down to
_get_dependencies_flags(), and many places had 'if depends' which is
False when 'depends' is an empty list and not only when it's None. There
is no reason for that argument to be optional, we always want to collect
dependencies.
John Ericson
Christoph Reiter
Federico Mena Quintero
Score_Under
Jussi Pakkanen
Ninja-Koala
John Ericson
Elliott Sales de Andrade
Thibault Saunier
Iñigo Martínez
Olivier Crête
TingPing
Ting-Wei Lan
Ross Burton
Marco Trevisan (Treviño)
Mathieu Duponchelle
Nirbheek Chauhan
Jon Turney
Xavier Claessens
Corentin Noël