On 2025-11-14, Maxim Cournoyer wrote:
> Noé Lopez <noe@noé.eu> writes:
>> Maxim Cournoyer <[email protected]> writes:
>>> Andreas Enge <[email protected]> writes:
>>>> the current policy is to not merge to master, but to only push on top.
>>>> This is not consistent with your proposal, Maxim. It is not set in
>>>> stone, but we should also not change the policy without thorough
>>>> discussion (and maybe a GCD). My suggestion would be to follow the
>>>> agreed-upon release process for now, and then in the subsequent
>>>> discussion period bring up problems (your concern may or may not turn
>>>> out to be a big blocker) and modify the release process accordingly.
>>>> But let us not change the process in the middle of our first release
>>>> since years ago.
>>>
>>> We're only discussing changing a detail that would likely make it easier
>>> for everyone. My fear here is that branches being worked on such as the
>>> Python branches, Qt, etc. are held back for a couple weeks (if
>>> everything goes as plan) "just because".  Working on a release branch
>>> and not blocking built branches merges to master while doing so appears
>>> win-win to me.  It's also simpler to manage from a CI/QA point of view.
>>>
>>
>> They are not held back, they are just merged to a different branch.
>
> Good point, I had overlooked that. Still, it assumes every of the 47
> current Guix committers will have read this and not make any mistake and
> push their large branch to master after it's been built by CI and QA.
>
> Having a distinct release branch would avoid the extra communication
> needed here and prevent the unavoidable mistakes.

This seems to be the most compelling point ... I could easily see folks
even accidentally pushing something inappropriate to master, even if
they know that master was in the release preparation phase, as that is
the usual workflow and there is nothing (to my knowledge) technically to
prevent it from happening...

I do not have a specific count, but I feel like there were a handful of
massive rebuilds accidentally pushed to master a few times over the last
year (some reverted, some not?), as a similar example where the policy
does not stop people from these sorts of (hopefully innocent)
mistakes. Seems like it could be the same for pushing to master during a
release freeze of any form.

So, while I very much welcomed the idea that the master branch would
maybe occasionally slow down a bit during release preparation (which
feels fine and cozy to me, coming from Debian)... the practicality of
actually "enforcing" it may be more difficult than initially realized.


At the danger of slipping towards off-topic, it really makes me wish
guix had "guix pull" default to some branch other than master, so that
master could just get all the latest changes, but only gets manually
fast-forward merged into whatever branch "guix pull" pulls by default
after some checks have passed, packages built, etc. ...


live well,
  vagrant

Attachment: signature.asc
Description: PGP signature

  • Toolchain and tra... Development of GNU Guix and the GNU System distribution.
    • Re: Toolchai... Maxim Cournoyer
      • Re: Tool... Development of GNU Guix and the GNU System distribution.
        • Re: ... Maxim Cournoyer
          • ... Development of GNU Guix and the GNU System distribution.
            • ... Andreas Enge
              • ... Maxim Cournoyer
                • ... Development of GNU Guix and the GNU System distribution.
                • ... Maxim Cournoyer
                • ... Vagrant Cascadian
                • ... Tomas Volf
                • ... Maxim Cournoyer
      • Re: Tool... Rutherther

Reply via email to