In commit e5a6283c4c, this function was
reorganized to assign value -> newvalue instead of overwriting newvalue,
but the error message case wasn't updated to match. As a result,
actually hitting the error would report an even more errory error, i.e.
a traceback.
(Actually hitting the error requires passing an array option as a python
object that isn't a list or a string. This is impossible to do from the
command line or meson_options.txt, but can be done with builtin options.)
They claimed that all of these functions accepted any posargs or varargs
that install scripts supported. This was inconsistent with both the
types that meson currently allowed, and the types that we documented in
refman 1.0 *should* be allowed.
Take the opportuninty to be clear as refman 1.0 never was, about the
difference between types supported in the first posarg, and the ypes
supported in succeeding varargs.
In commit c239ce31f5 support was added to
these functions to accept various non-string types.
Despite the commit/PR documenting that only add_install_script is
permitted to accept built files, the actual check parameter was set, for
all three, to "True" (so the function was never invoked with False at
all). This meant that actually attempting to use the allowed types would
fail at postconf or dist, with python tracebacks in the former case and
"Failed to run dist script" in the latter case.
This was partially ameliorated in commit
6c5bfd4c24 which added typed_pos_args, but
unfortunately those typed_pos_args were all over the place.
For postconf:
- They banned external programs as additional args (which should be allowed)
- They banned built executables (good)
- They allowed custom targets as additional args (bad)
For dist:
- they allowed external programs (good)
- they allowed built executables, but only as the first argument (bad, also ???)
- they allowed custom targets, but only as additional arguments (bad, also ???)
Fix this all to only allow the same argument types for both the script
argument and the script-args arguments. That type is known at configure
time and restricted to source files, configured files, and found
programs.
In commit 2c0eaf5c4f support was added for
install scripts to accept found programs, built executables, or custom targets.
In commit c239ce31f5, this was extended to
dist and postconf scripts too (although it was documented that those
should not accept targets that are built by ninja).
Despite the commit/PR claiming that all of these should always accept
files and configured files, this was only true for arguments other than
the first, until commit f808c955ea.
In amongst all this, FeatureNew checks were never registered for the
first argument, only for additional arguments, until late in the game
with the addition of FeatureNew checks for File objects.
Fix this in part by moving the 3 different File checks into one, inside the
function that processes the first script, and make that function check
for FeatureNew on anything else too.
multiple versions of python are packaged by cygwin, and the default
python is auto-selected as the latest one via a Debian-like alternatives
system.
This recently broke because dblatex is built against 3.9, causing it to
be installed too and resulting in multiple (inconsistent) versions of
python being installed, and `python3` pointing to the one we don't have
devel packages for and isn't even the default version.
Fix this by pointing back to the intended python.
Removed errant "type: ignore".
Fixed issue with "fetch" call. This issue was the following:
Dict::get() and Dict::pop() have the following signature:
T.Callable[[_T, _U], _U | None] OR T.Callable[[_T], _U | None]
Note how the return type is _U here. When the fetch() function was
actually being called, it had the following signature:
T.Callable[[_T, T.List[_U]], T.Union[T.List[_U], _U]]
This is incompatible with the previous definitions. The solution is
simply to move where the default value is introduced if fetch() produces
None.
pkgconf has a bug on MSYS2 due to which prefixes with spaces are not
handled correctly if the library has a Requires: on another library
and both have prefixes with spaces in them.
See: https://github.com/pkgconf/pkgconf/issues/238
So move the unit test to libanswer.pc instead of libfoo.pc till that
is fixed.
Since we scan all dependencies for build-only RPATHs now (which are
removed on install), we must take care not to add build-only RPATHs
pointing to directories that dependencies explicitly add -Wl,-rpath
link args for, otherwise the paths will get wiped on install.
Caught by LinuxlikeTests::test_usage_pkgconfig_prefixes
If a pkg-config dependency has multiple libraries in it, which is the
most common case when it has a Requires: directive, or when it has
multiple -l args in Libs: (rare), then we don't add -Wl,-rpath
directives to it when linking.
The existing test wasn't catching it because it was linking to
a pkgconfig file with a single library in it. Update the test to
demonstrate this.
This function was originally added for shared libraries in the source
directory, which explains the name:
https://github.com/mesonbuild/meson/pull/2397
However, since now it is also used for linking to *all* non-system
shared libraries that we link to with absolute paths:
https://github.com/mesonbuild/meson/pull/3092
But that PR is incomplete / wrong, because only adding RPATHs for
dependencies that specify a single library, which is simply
inconsistent. Things will work for some dependencies and not work for
others, with no logical reason for it.
We should add RPATHs for *all* libraries. There are no special length
limits for RPATHs that I can find.
For ELF, DT_RPATH or DT_RUNPATH are used, which are just stored in
a string table (DT_STRTAB). The maximum length is only a problem when
editing pre-existing tags.
For Mach-O, each RPATH is stored in a separate LC_RPATH entry so there
are no length issues there either.
Fixes https://github.com/mesonbuild/meson/issues/9543
Fixes https://github.com/mesonbuild/meson/issues/4372
When generating pkgconfig file for a library that links to an
uninstalled static library built by custom_target() Meson was crashing
when trying to access some attributes that does not exist on that class.
Also fix is_internal() implementation, it only really make sense on a
CustomTargetIndex or if CustomTarget has only a single output.
Add ability to mutate a target's `extra_files` list through the
rewriter.
The logic is copied from sources add/rm, but changes the `extra_files`
kwarg instead of the sources positional argument.
Has additional logic to handle creating the `extra_files` list if it
doesn't exist.
In 1fb6c93947 the intl dep gained support for static linking
which also forwarded this property to the iconv sub dependency.
The refactoring in 214d03568f lost this change, which results in
iconv getting linked dynamically again.
Forward static again to fix this.
These are actually just flags, they don't take any arguments (except
prefix, which was already handled correctly), and as such their
arguments should be booleans, not strings, and they should default to
False.
Currently there is a try/except around the function that detects and
rejects this, which instead of rejecting it, spawns a warning and
continue.
This warning exists because of 'test cases/vala/9 gir/' which passes a
vala generated output that isn't a return value (!!!) using string
joining with the meson.current_build_dir() function (also !!!) because
we officially document this (!!! for a third time) as the only way to
make a vala shared library generate a typelib with a custom_command from
the automatically generated gir:
https://mesonbuild.com/Vala.html#gobject-introspection-and-language-bindings
In #3061 we converted strings to Files, but only if none of them were
this vala hack. Due to the precise implementation, we also failed to
convert strings to Files if any other error occurred, but since we only
want to ignore errors for generated vala outputs, tighten that check and
specifically call out generated files in the warning.
Fixes#8635
There is a change here, in that packages has error messaging for using
IncludeDirs objects in the packages argument, but it never worked, so
the message was useless.
There are thee arguments that are passed directly to CustomTarget which
are not in the permittedKwargs: depends, depend_files, and build_always.
The first two are obviously generically useful, as they allow creating
correct ordering. The latter, not so much. Since it was an error to pass
it, we'll just delete it.
Use a proper install option for this. Now `install_<type>` can directly
override `install` instead of passing a boolean to the string kwarg
`install_dir_<type>`.
Eli Schwartz
Tristan Partin
Nirbheek Chauhan
Xavier Claessens
Dylan Baker
Matthias Klumpp
Daniel Mensinger
Rafael Silva
Ben Boeckel
Celeste Wouters
Christoph Reiter
Jussi Pakkanen
William Toohey
Sahnvour