RFC: The Binutils and LLM generated patches

[Matt Rice]

On Fri, Oct 3, 2025 at 10:46 AM Nick Clifton <nickc@redhat.com> wrote:
>
> Hi Jose,
>
> >>    The proposed text also invents two new tags for LLM generated
> >>    contributions:  Generated-By:  and  LLM-Query:  I just pulled these
> >>    out of thin air, so feel to suggest better ones, or tell me that they
> >>    are pointless.
> >
> > If these tags are to be used only for trivial/non-copyrightable changes,
> > as I think/hope the policy establishes, I personally see no point in
> > having these tags.  Trivial changes are trivial regardless of how they
> > are composed, so who cares whether a LLM was involved in it?
>
> Ah - because I am hoping to establish a precedent.  If it becomes accepted
> that LLM generated patches must identify themselves, and use the tags I
> have suggested then when/if we ever start accepting non-trivial LLM patches
> we will already have the framework in place for how those patches should be
> annotated.
>

I feel like Generated-By: is probably also useful for things like code
formatters
and static analysis tools that are not LLM, there recently has been
discussion of
adding a .clang-format file to the top level, (but definitely no
requirement to run it).

But from a review perspective it seems that it would be useful to know
if some change
was generated by a code formatter or linter/static analysis tool.  So
I think the tag
could be useful even outside of LLM specific usage.

> > If I create a patch that removes trailing whitespaces in a file using
> > sed, or a patch that renames a variable in the whole project using
> > coccinelle, Generated-By would also apply to them.  Still, we don't feel
> > the need to do so.
>
> Well I cannot force anyone to use these tags, so in the end it will be up
> the contributor to decide if they can be bothered.  I agree that for
> trivial patches it is not really going to matter, but I do like the idea
> of setting up a precedent for future non-trivial submissions.
>
> Cheers
>    Nick
>
>