Currently we do some crazy hackery where we add extra properties to a
Popen object and return that. That's crazy. Especially since some of our
hackery is to delete attributes off of the Popen we don't want. Instead,
let's just have a discrete type that has exactly the properties we want.
We're using these now, so having some error checking to make sure we
don't have paths were we're trying to instantiate an abstract class
would be good.
This catches some very real errors.
The one in scalapack is pretty silly actually, it's failing to figure
out that the exploded list is at least two arguments. However, the code
is actually clearer by not using a list and exploding it, so I've done
that and pylint is happy too.
1. Like with gcc's `ld`, also use the `group_start` code to create a
`--start-group`/`--end-group`
2. xc16 tricked into believing the 'link_whole' was about `--*-group`,
but it should use gcc's `--whole-archive` instead.
3. Not clear what the get_lib_prefix should really do, but for picolibc
it seems I want just `''`.
The problem with picolibc was that the `-l` would be prefixed to a lib
like `picolib/libm/libm.a`. Though of course the `-l` would be necessary
for just a plain `m` (that's what I assumed this would be used for).
I think this might need some clarification from the meson devs ;-)
It was done to include them in `meson subprojects foreach` without
--types argument, but it's better to special case missing --types and
include wraps that have type=None too. It was a bad idea because that
was messing them in `meson subprojects update`, now they are ignored by
that command.
If revision is a tag that does not exist locally, `git fetch origin
<revision>` won't create it and checkout will fail. Using --refmap
ensures that references exists locally.
This is a test for https://github.com/mesonbuild/meson/pull/7021, to
verify that `link.exe` uses the correct architecture when targeting
ARM64. Can be extended to other cross targets later.
There are two problems here. The first is that we're doing manual monkey
patching which is fragile and verbose, when unittest.mock is designed
specifically to solve this exact problem. The second is that we're
monkey patching os.environ at two different levels with the same
information. So let's only do it once.
When building with vs2019 (not ninja), a path length error will be thrown
if the path to a resource file is even remotely deep within the tree.
This is largely because the target name includes the string "Windows
resource for file 'full path'", which is then expanded twice (once for
the .vcxproj itself, and once for IntDir) and added to the full path.
When combined with the tiny path limits on Windows, it is easy to exceed
path limits.
This error is largely avoided by the ninja back-end. Unlike the
vs back-end, the ninja back-end does not use target.get_id() as part of
the project file path, nor does it use target.get_id() as part of
get_target_private_dir().
Example error:
error MSB4184: The expression "[MSBuild]::NormalizePath(
C:\src\mesonbuild\Misc\FreeRDP-master\client\X11\xfreerdp\xfreerdp,
f3f7317@@Windows resource for file
'Misc_FreeRDP-master_client_X11_xfreerdp_xfreerdp_xfreerdp.rc'@cus\,
f3f7317@@Windows resource for file
'Misc_FreeRDP-master_client_X11_xfreerdp_xfreerdp_xfreerdp.rc'@cus.
vcxproj.CopyComplete)" cannot be evaluated. Path:
C:\src\mesonbuild\Misc\FreeRDP-master\client\X11\xfreerdp\xfreerdp\f3f7317
@@Windows resource for file
'Misc_FreeRDP-master_client_X11_xfreerdp_xfreerdp_xfreerdp.rc'@cus\f3f7317
@@Windows resource for file
'Misc_FreeRDP-master_client_X11_xfreerdp_xfreerdp_xfreerdp.rc'@cus.
vcxproj.CopyComplete exceeds the OS max path limit.
The fully qualified file name must be less than 260 characters.
If the architectures are taken from the output of "clang-cl --version",
we need to convert these names into names that the MSVC tools accept
as the -machine: parameter.
Originally I had this idea that you'd be able to pass the id in to be
able to deduplicate some cases (like ld.gold and ld.bfd). That went away
because it ended up being really un-dry, but this id per instance
remained. Getting rid of it allows us to get rid of a bunch of otherwise
useless super calls, which makes adding type annotations easier.
Dylan Baker
Nirbheek Chauhan
Jussi Pakkanen
Sebastian Meyer
Martin Storsjö
Xavier Claessens
Nirbheek Chauhan
Peter Harris
Marco Mastropaolo