Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
References: <20230508124701.443ba331.gary@garyguo.net>
 <20230509080700.1608146-1-aliceryhl@google.com>
 <2023050937-scouring-truck-d7c0@gregkh>
 <CANiq72=g=CnEQTkc50qozUdOBZECr7Jr2NCRfGFaKh751i6Sww@mail.gmail.com>
User-agent: mu4e 1.10.3; emacs 28.2.50
From:   Andreas Hindborg <nmi@metaspace.dk>
To:     Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
Cc:     Greg KH <gregkh@linuxfoundation.org>,
        Alice Ryhl <aliceryhl@google.com>, gary@garyguo.net,
        alex.gaynor@gmail.com, benno.lossin@proton.me,
        bjorn3_gh@protonmail.com, boqun.feng@gmail.com,
        linux-kernel@vger.kernel.org, ojeda@kernel.org,
        patches@lists.linux.dev, rust-for-linux@vger.kernel.org,
        wedsonaf@gmail.com
Subject: Re: [PATCH v2] rust: error: add missing error codes
Date:   Mon, 15 May 2023 20:07:12 +0200
In-reply-to: <CANiq72=g=CnEQTkc50qozUdOBZECr7Jr2NCRfGFaKh751i6Sww@mail.gmail.com>
Message-ID: <875y8teab3.fsf@metaspace.dk>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk


Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> writes:

> On Tue, May 9, 2023 at 10:46=E2=80=AFAM Greg KH <gregkh@linuxfoundation.o=
rg> wrote:
>>
>> It depends on which arch you are building for.  That's why we have
>> per-platform errno.h files, the values are different for different ones.
>> So you need to handle them all properly somehow.  How is rust going to
>> handle per-arch stuff like this?
>
> We can do conditional compilation in the same file, possibly with a
> Rust macro which takes a nice table that shows all arches at once.
>
> We can also split into files like C and move it to each `arch/`, there
> are a couple of approaches for this. This is best for `MAINTAINERS`,
> although these headers almost never change, so it is not a big
> advantage.
>
> We could also automatically do everything based on the C headers, too.
> Back then it felt to me like too much complexity for little gain,
> given those C headers almost never change, but now it may be worth it.
> Or, instead, having a test that verifies they are the same instead,
> and that way we don't introduce complexity for the build itself.
>
> Alice only needs `ERESTARTSYS` so far, as far as I understand, so
> perhaps it is simplest to only add the rest of the non-generic ones
> for the moment; and gather opinions on the approaches above meanwhile.

Let's add the ones we need for now. When we need target specific error
codes we can have a `mod` for each arch, gate behind the target
feature and conditionally reexport them.

BR Andreas