Re: [PATCH v4 2/4] ynl_gen: generate Rust files from yaml files

Re: [PATCH v4 2/4] ynl_gen: generate Rust files from yaml files

[Jakub Kicinski]

On Mon, 04 May 2026 09:04:55 +0000 Alice Ryhl wrote:
>  tools/net/ynl/pyynl/ynl_gen_c.py | 139 ++++++++++++++++++++++++++++++++++++++-

No. Rust. In. This. File.

Just commit the artifacts. I truly hope that this is the only Netlink
family we will have in Rust.
-- 
pw-bot: cr

Re: [PATCH v4 2/4] ynl_gen: generate Rust files from yaml files

[Alice Ryhl]

On Mon, May 04, 2026 at 04:58:58PM -0700, Jakub Kicinski wrote:
> On Mon, 04 May 2026 09:04:55 +0000 Alice Ryhl wrote:
> >  tools/net/ynl/pyynl/ynl_gen_c.py | 139 ++++++++++++++++++++++++++++++++++++++-
> 
> No. Rust. In. This. File.
> 
> Just commit the artifacts. I truly hope that this is the only Netlink
> family we will have in Rust.

There's no reason to react like this. I have not ignored your concern.
Last time we discussed this, the discussion ended on splitting the file
into ynl_gen_c.py and ynl_gen_rust.py, which you did not reply to, and I
actually spent some time working on that. However, I felt the change was
non-trivial and I wanted to discuss whether that was the correct way
forward before spending more time on it. Therefore, I kept this patch
as-is for now and noted why it was non-trivial (sharing of CodeWriter)
in the commit message, until we could discuss further.

I think you are probably right that just comitting the artifacts is the
simplest way forward for now. Especially since Donald is apparently
working on splitting up the file for strace [1]. On the off-chance that
a second Netlink family is ever added, hopefully Donald's work has
already completed and we can easily add this support in a new file when
the time comes.

I guess another way forward is to commit a copy of the python script
with the Rust support to drivers/android/binder/ and I can run it
manually if the Binder yaml file is ever updated.

Alice

[1]: https://lore.kernel.org/r/CAD4GDZxGo6p9A07rKAW4MZNdFOYrMVzjkYNjwTqKfFcqKkzU1g@mail.gmail.com

Re: [PATCH v4 2/4] ynl_gen: generate Rust files from yaml files

[Jakub Kicinski]

On Tue, 5 May 2026 09:10:17 +0000 Alice Ryhl wrote:
> On Mon, May 04, 2026 at 04:58:58PM -0700, Jakub Kicinski wrote:
> > On Mon, 04 May 2026 09:04:55 +0000 Alice Ryhl wrote:  
> > >  tools/net/ynl/pyynl/ynl_gen_c.py | 139 ++++++++++++++++++++++++++++++++++++++-  
> > 
> > No. Rust. In. This. File.
> > 
> > Just commit the artifacts. I truly hope that this is the only Netlink
> > family we will have in Rust.  
> 
> There's no reason to react like this. I have not ignored your concern.
> Last time we discussed this, the discussion ended on splitting the file
> into ynl_gen_c.py and ynl_gen_rust.py, which you did not reply to, and I
> actually spent some time working on that. However, I felt the change was
> non-trivial and I wanted to discuss whether that was the correct way
> forward before spending more time on it. Therefore, I kept this patch
> as-is for now and noted why it was non-trivial (sharing of CodeWriter)
> in the commit message, until we could discuss further.
> 
> I think you are probably right that just comitting the artifacts is the
> simplest way forward for now. Especially since Donald is apparently
> working on splitting up the file for strace [1]. On the off-chance that
> a second Netlink family is ever added, hopefully Donald's work has
> already completed and we can easily add this support in a new file when
> the time comes.
> 
> I guess another way forward is to commit a copy of the python script
> with the Rust support to drivers/android/binder/ and I can run it
> manually if the Binder yaml file is ever updated.

Could you _please_ do what I'm asking you to do instead inventing your
own solutions. Just commit the generated files and leave the script out.
We lived without Netlink code gen for 30 years.

Re: [PATCH v4 2/4] ynl_gen: generate Rust files from yaml files

[Alice Ryhl]

On Tue, May 05, 2026 at 05:16:37PM -0700, Jakub Kicinski wrote:
> On Tue, 5 May 2026 09:10:17 +0000 Alice Ryhl wrote:
> > On Mon, May 04, 2026 at 04:58:58PM -0700, Jakub Kicinski wrote:
> > > On Mon, 04 May 2026 09:04:55 +0000 Alice Ryhl wrote:  
> > > >  tools/net/ynl/pyynl/ynl_gen_c.py | 139 ++++++++++++++++++++++++++++++++++++++-  
> > > 
> > > No. Rust. In. This. File.
> > > 
> > > Just commit the artifacts. I truly hope that this is the only Netlink
> > > family we will have in Rust.  
> > 
> > There's no reason to react like this. I have not ignored your concern.
> > Last time we discussed this, the discussion ended on splitting the file
> > into ynl_gen_c.py and ynl_gen_rust.py, which you did not reply to, and I
> > actually spent some time working on that. However, I felt the change was
> > non-trivial and I wanted to discuss whether that was the correct way
> > forward before spending more time on it. Therefore, I kept this patch
> > as-is for now and noted why it was non-trivial (sharing of CodeWriter)
> > in the commit message, until we could discuss further.
> > 
> > I think you are probably right that just comitting the artifacts is the
> > simplest way forward for now. Especially since Donald is apparently
> > working on splitting up the file for strace [1]. On the off-chance that
> > a second Netlink family is ever added, hopefully Donald's work has
> > already completed and we can easily add this support in a new file when
> > the time comes.
> > 
> > I guess another way forward is to commit a copy of the python script
> > with the Rust support to drivers/android/binder/ and I can run it
> > manually if the Binder yaml file is ever updated.
> 
> Could you _please_ do what I'm asking you to do instead inventing your
> own solutions. Just commit the generated files and leave the script out.
> We lived without Netlink code gen for 30 years.

I will drop the patch in the next version.

Alice