Cpp23 import std attempt2

[mccakit]

This PR adds import std support for clang

[mathstuf]

I see that it is using compile_commands.json for clang-scan-deps. Are generated sources that are created during the build supported here? If not, is that wanted? If so, can docs about that be added?

I've only done a skim rather than a deep dive so far. One thing that would help would be test cases.

[mccakit]

This works

project('hello_modules', 'cpp',
  default_options: ['cpp_std=c++26'])

gen = custom_target('gen_module',
  output: 'node_generated.cpp',
  command: ['python3', files('gen_module.py'), '@OUTPUT@'])

exe = executable('hello',
  sources: ['main.cpp', 'node_a.cc', 'node_b.cpp', 'node_c.cxx', 'node_d.cc', gen])

You can test it yourself if you want

hello_modules.zip

[mccakit]

Currently, the only failing test cases are related to the Fortran and Rust modules. I would recommend removing that functionality from the build system, since those packages should be built using their native build systems — Rust with Cargo (pkgconfig packages by cargo-c), and Fortran with its respective toolchain.

That said, if anyone has a better solution for handling this, suggestions are welcome. I don’t have much experience with Fortran compilation, nor is it something I’m particularly focused on.

Once this is resolved, I would add the test project I shared in this discussion as a test case, and the remaining tests can be derived from the ongoing discussion.

[mccakit]

Last commit fixes fortran and rust module errors by allowing per target scanning for fortran.

[mccakit]

I'm having trouble fixing the CI for ubuntu rolling clang.
With the last commit I can run tests without any errors.
Some help would be appreciated

[mccakit]

I tried replicating the ubuntu clang ci, and tests passed. Only thing that was need was a symlink for finding clang-scan-deps. But I think that needs to be a variable on its own but currently I'm not planning on any new changes.

[jpakkane]

The pyright config file is still in this MR. All traces of it should be rebased away (i.e. removing the "Delete pyrightconfig.json" commit and everything related).

[mccakit]

The pyright config file is still in this MR. All traces of it should be rebased away (i.e. removing the "Delete pyrightconfig.json" commit and everything related).

Done

[mccakit]

ping @jpakkane

[mccakit]

ping @dcbaker @mathstuf

[dnicolodi]

The lint failures needs to be fixed. The test failure seem relevant to the changes to the PR and need to be at least investigated. The motivation to look at the PR before these things are fixed is very little.

[mccakit]

How am I supposed to know what needs to be fixed if there is no feedback?

I'm not a long time contributor to meson, reviewers should guide me a little or should close my PR to show their disinterest.

[dnicolodi]

How am I supposed to know what needs to be fixed if there is no feedback?

I just provided you feedback. However, I think the fact that the lint checks and the projects tests need to pass should be self evident. Resources are finite and most likely no one feels like spending the available resources reviewing code that is known to require more work before it is working order.

[mccakit]

Alright, lint errors are gone, I really need some feedback for other CI failures.

[mccakit]

I need some information on how the llvm detected with the macos CI, framework test always fail due to not finding llvm.

I'm not a Mac user, so I have no clue how to fix this, I tried editing llvm version detection in the source and installing a specific llvm version via homebrew but got the same result.

[mccakit]

@germandiagogomez hey, from the draft I can assume you are a Mac user.
Would you like to assist me in this PR?

I can't get the 2/3 macos workflows succeed, and I only have windows and linux machines available.

[mccakit]

Ok macos tests pass now, are there any tips on how to deal with ci logs?
I need to focus on one error at a time, but log is like 100k lines.

[mccakit]

with last commit ubuntu ci issues are gone, at least in my own container where I run the tests, I'll update the comment if CI fails.

However, generated module files no longer work.

[mccakit]

CI image failiures and msys/cygwin issues have nothing to do with this PR.
I can't fix cmake ones, that is beyond me. I have no clue how meson invokes cmake etc.
If I won't get any feedback on this, I'll turn the PR into a draft and let someone else pick it up in the future kinda like the og PR here, #14989