Re: Debian & bindgen issue

[Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>]

Hi NoisyCoil,

On Wed, Oct 30, 2024 at 10:28 PM NoisyCoil <noisycoil@disroot.org> wrote:
>
> I agree that we should probably upload a newer bindgen (especially so it
> receives enough testing before the upcoming freeze). In the meantime I
> opened a MR for the backport [1]. Miguel, do you have an estimate for

Thanks a lot!

> when this bug is going to start breaking kernel builds? If it's not
> urgent we could skip the backport and just fix it with the update. In
> any case, yes, please do file a bug report so that we can track this
> (e.g. I'll add it to [1]'s d/changelog if we decide to go forward with it).

Matthias replied to me in the meantime (but the Cc got dropped), and I
replied back with some more details (including the timeline) --
copying it here:

  From a quick check, the commit seems to be able to be cherry-picked
  cleanly in bindgen 0.66.1. It is just a single line change (a function
  name change in a call).

  I also tested downloading the Sid package sources (rust-bindgen and
  rust-bindgen-cli), building them, checking that the issue is there
  when building the kernel, then applied the single line change to the
  sources, building it again, and checked that the issue disappears
  building the kernel.

  Of course, that does not mean it wouldn't break something else in Debian.

  We expect the code to hit mainline in Linux v6.13 (the merge window
  for that one will be in a few weeks); perhaps v6.14. So the
  consequence of not doing anything would be that Debian users (kernel
  developers, mostly, I assume) would essentially need to install a
  custom/newer bindgen, possibly via `rustup`.

I will open the bug report -- happy to test the package with a kernel
build too if that helps.

Thanks!

Cheers,
Miguel