bug#81520: Add pre-receive hook to Savannah to reject LLM-encumbered commits

> Date: Fri, 31 Jul 2026 18:36:58 +0300
> Cc: [email protected], [email protected], [email protected]
> From: Dmitry Gutov <[email protected]>
> 
> On 31/07/2026 18:26, Eli Zaretskii wrote:
> >> Date: Fri, 31 Jul 2026 16:08:45 +0300
> >> Cc: [email protected], [email protected]
> >> From: Dmitry Gutov <[email protected]>
> >>
> >> Hi Eli,
> >>
> >> On 31/07/2026 14:45, Eli Zaretskii wrote:
> >>> My point is that when I have several commits, and then pushing them
> >>> all fails because one of them, going back and redoing the one that
> >>> caused the fail is a terrible nuisance with Git.  So I want the
> >>> rejection happen when I commit, so I could fix it right away.
> >>
> >> If the pre-push hook only checks for LLM attributions, and does that
> >> with decent accuracy, would your scenario above ever happen?
> > 
> > I don't see why not.  I explained up-thread how it could happen.
> 
> I mean... if that does happen, wouldn't you want it to be stopped at the 
> "git push" stage?

No, I want to be stopped at the commit stage, where fixing this is
easier.

> If you agree, then the question might be whether both pre-push and 
> pre-commit checks are needed.

I already said that I'm okay with having both, if needed.

> >> I don't imagine you would have commits with Assisted-By header in the
> >> local checkout of Emacs, with any regularity.
> > 
> > I think you forget that I frequently push commits of others, and also
> > merge from the release branch to master, and then push.
> 
> Hopefully the release branch isn't going to receive problematic 
> messages, so merging from it won't create such difficulty.

There's the --no-validate switch to Git, which bypasses hooks.  So
yes, such commit could be on a branch.