[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CANiq72knS9dGTh1uEQzYvmZuFz6fE8tNhZ8JEuyKFHYR=Kh8sQ@mail.gmail.com>
Date: Sat, 10 Aug 2024 13:36:02 +0200
From: Miguel Ojeda <miguel.ojeda.sandonis@...il.com>
To: Masahiro Yamada <masahiroy@...nel.org>
Cc: Miguel Ojeda <ojeda@...nel.org>, Alex Gaynor <alex.gaynor@...il.com>,
Wedson Almeida Filho <wedsonaf@...il.com>, Boqun Feng <boqun.feng@...il.com>, Gary Guo <gary@...yguo.net>,
Björn Roy Baron <bjorn3_gh@...tonmail.com>,
Benno Lossin <benno.lossin@...ton.me>, Andreas Hindborg <a.hindborg@...sung.com>,
Alice Ryhl <aliceryhl@...gle.com>, rust-for-linux@...r.kernel.org,
Nathan Chancellor <nathan@...nel.org>, Nicolas Schier <nicolas@...sle.eu>, linux-kbuild@...r.kernel.org,
linux-kernel@...r.kernel.org, patches@...ts.linux.dev
Subject: Re: [PATCH 5/6] kbuild: rust: replace proc macros dependency on
`core.o` with the version text
On Fri, Aug 9, 2024 at 7:31 PM Masahiro Yamada <masahiroy@...nel.org> wrote:
>
> The touched file, include/config/*, is an implementation detail
> in Kconfig and fixdep.
>
> Rather, I'd like to put the string "CONFIG_RUST_VERSION_TEXT"
> in the comment of the source file.
>
> This is the idea adopted in include/linux/compiler-version.h
+1, I did it this way to follow the same pattern to the previous patch
on `core.o`, since that one needed another approach, but I am happy to
change it.
> I do not know how to do it for rust/core.o because there is no in-tree
> source file.
>
> But, can we add rust/core.rs, from which
> rustlib/src/rust/library/core/src/macros/mod.rs is imported?
That is an interesting idea... :)
Hmm... I think `core` is a bit too special for that, since we need
attributes on the crate root and it is the one defining things like
`include!`. Even if we generated the file on the fly to mimic the
original, we would still need to handle the paths for each module or
recreate a directory hierarchy or similar.
We are also in talks with the Rust project to figure out building
`core` in a stable way, so the details around it may change too, e.g.
probably something like passing a flag like `-Cbuild-std=core`. They
may also want to figure out the right input file path themselves, i.e.
rather than having to provide one from our side.
So I think a hack may not be worth it compared to depending on the
implementation detail, and may be a bit brittle until we know the
final details of that.
Thanks!
Cheers,
Miguel
Powered by blists - more mailing lists