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

Jim Porter [18/Aug 11:14am -07] wrote:
> Alternately, I could write a test script that initializes an all-new
> Git repo and runs through some cases.

This makes sense to me.

>> Maybe we could match 'claude' case-sensitively?
>
> But that could hit people whose email addresses contain the substring 
> "claude". We do have existing contributions that would have been flagged 
> with a check like this, from Clément Pit-Claudel.

Fair enough.

> Attached is a script you can use to run the LLM check on historical 
> commits (it defaults to the last 1000, but pass any number you like as a 
> command line arg). I've run this on the last 15,000 commits and saw no 
> errors.

Thanks, that's reassuring.

> +** LLM usage policies
> +
> +The FSF is currently reviewing the legal and ethical implications of
> +large language models on free software.  In the meantime, the GNU
> +project doesn't allow LLM-generated outputs in contributions.  This

s/outputs/output/

> +applies to all code, documentation, test data, and commit messages.
> +
> +This is enforced by a git hook that runs when committing or pushing

It's enforced by the maintainers, so s/enforced/checked/.

> +AUTHOR=$(git var GIT_AUTHOR_IDENT | sed -n 's/^\(.*>\).*$/\1/p')
> +COMMITTER=$(git var GIT_COMMITTER_IDENT | sed -n 's/^\(.*>\).*$/\1/p')

It would be good to have a comment explaining these transformations.

Otherwise, I didn't read very carefully but this seems much better.

-- 
Sean Whitton