[RFC] glibc AI policy. Round 3. Adopt the GCC AI policy.
Carlos O'Donell
carlos@redhat.com
Tue Nov 3 23:08:45 GMT 2026
On 10/28/26 1:00 PM, DJ Delorie wrote:
> On Wed, Oct 28, 2026 at 5:30 AM Jonathan Wakely <jwakely@redhat.com> wrote:
>> Why? I don't think "Reviewed-by: LLM" is useful.
>
> I think "Reviewed-by: LLM" is dangerous. I assume there are scripts
> out there that assume any RB tag means the patch is approved for
> commit, which is not true if it's an LLM.
>
> IMHO LLM reviews should be fail-only or advisory, but never count as
> approval.
Slightly OT, but...
I just wanted to point out that Reviewed-by: in glibc, while often
followed by a commit of that patch, isn't what is required to commit
code :-)
We require consensus on the technical matters at hand among the
developers of the project.
A review is a pretty good signal of consensus though, but not the
ultimate goal. The goal is consensus as a mechanism for improving
the implementation of a core library.
Can an LLM contribute to consensus? Yes, I think like any tool,
grep, sed, or gcc, can let me elucidate some truth that I will
use on furthering consensus. But consensus is ultimately something
that humans create together.
--
Cheers,
Carlos.
More information about the Libc-alpha
mailing list