Feedback-ID: iad51458e:Fastmail
Date:   Sun, 26 Feb 2023 14:13:10 -0800
From:   Boqun Feng <boqun.feng@gmail.com>
To:     Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>
Cc:     Gary Guo <gary@garyguo.net>, Asahi Lina <lina@asahilina.net>,
        Miguel Ojeda <ojeda@kernel.org>,
        Alex Gaynor <alex.gaynor@gmail.com>,
        Wedson Almeida Filho <wedsonaf@gmail.com>,
        =?iso-8859-1?Q?Bj=F6rn?= Roy Baron <bjorn3_gh@protonmail.com>,
        Sven Van Asbroeck <thesven73@gmail.com>,
        Fox Chen <foxhlchen@gmail.com>, rust-for-linux@vger.kernel.org,
        linux-kernel@vger.kernel.org, asahi@lists.linux.dev
Subject: Re: [PATCH 5/5] rust: error: Add from_kernel_result!() macro
Message-ID: <Y/vZdv6IVmrcP7wR@boqun-archlinux>
References: <20230224-rust-error-v1-0-f8f9a9a87303@asahilina.net>
 <20230224-rust-error-v1-5-f8f9a9a87303@asahilina.net>
 <Y/lOlcSpc+d9ytq/@boqun-archlinux>
 <20230225222340.34d749ee.gary@garyguo.net>
 <Y/rCU1S+GDgIojNf@boqun-archlinux>
 <20230226133606.583cd1d8.gary@garyguo.net>
 <Y/uiAfZnfbbMrQuD@boqun-archlinux>
 <CANiq72kTiHF76T0AycM43qj4rUgQpHzBqXujdvk+H2qoDz22AQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <CANiq72kTiHF76T0AycM43qj4rUgQpHzBqXujdvk+H2qoDz22AQ@mail.gmail.com>
Precedence: bulk

On Sun, Feb 26, 2023 at 09:59:25PM +0100, Miguel Ojeda wrote:
> On Sun, Feb 26, 2023 at 7:17 PM Boqun Feng <boqun.feng@gmail.com> wrote:
> >
> > My preference to function instead of macro here is because I want to
> > avoid the extra level of abstraction and make things explict, so that
> > users and reviewers can understand the API behavior solely based on
> > Rust's types, functions and closures: they are simpler than macros, at
> > least to me ;-)
> 
> There is one extra problem with the macro: `rustfmt` does not format
> the contents if called with braces (as we currently do).

Interesting, sounds like a missing feature in `rustfmt` or maybe we
don't use the correct config ;-)

> 
> So when I was cleaning some things up for v8, one of the things I did
> was run manually `rustfmt` on the blocks by removing the macro
> invocation, in commit 77a1a8c952e1 ("rust: kernel: apply `rustfmt` to
> `from_kernel_result!` blocks").
> 
> Having said that, it does format it when called with parenthesis
> wrapping the block, so we could do that if we end up with the macro.
> 
> > First, I think the macro version here is just a poor-man's try block, in
> > other words, I'd expect explicit use of try blocks intead of
> > `from_kernel_result` when the feature is ready. If that's the case, we
> > need to change the use sites anyway.
> 
> Yeah, if we eventually get a better language feature that fits well,
> then we should use it.
> 
> > Do both implementation share the same behavior?
> 
> Yeah, a `return` will return to the outer caller in the case of a
> `try` block, while it returns to the closure (macro) in the other
> case. Or do you mean something else?
> 

"Yeah" means they have different behaviors, right? ;-)

Thanks for confirming and I think you get it, but just in case for
others reading this: if we use the macro way to implement
`from_kernel_result` as in this patch:

	macro_rules! from_kernel_result {
	    ($($tt:tt)*) => {{
		$crate::error::from_kernel_result_helper((|| {
		    $($tt)*
		})())
	    }};
	}

and later we re-implement with try blocks:

	macro_rules! from_kernel_result {
	    ($($tt:tt)*) => {{
		$crate::error::from_kernel_result_helper(try {
		    $($tt)*
		})
	    }};
	}

the `from_kernel_result` semantics will get changed on the `return`
statement inside the macro blocks.

And this is another reason why we want to avoid use macros here. Code
example as below:

	https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=58ea8b95cdfd6b053561052853b0ac00

`foo_v1` and `foo_v3` has the exact same function body, but behave
differently.

> In that case, I think one could use use a labeled block to `break`
> out, not sure if `try` blocks will allow an easier way.
> 
> We have a case of such a `return` within the closure at `rust/rust` in
> `file.rs`:

Thanks for finding an example! Means we did use return.

For this particular API, I'd say function right now, `try` blocks if
avaiable.

Regards,
Boqun

> 
>     from_kernel_result! {
>         let off = match whence as u32 {
>             bindings::SEEK_SET => SeekFrom::Start(offset.try_into()?),
>             bindings::SEEK_CUR => SeekFrom::Current(offset),
>             bindings::SEEK_END => SeekFrom::End(offset),
>             _ => return Err(EINVAL),
>         };
>         ...
>         Ok(off as bindings::loff_t)
>     }
> 
> Cheers,
> Miguel