Re: Sashiko review feedback (was Re: [PATCH 2/3] soc: renesas: Add Renesas R-Car MFIS driver)

["Theodore Tso" <tytso@mit.edu>]

On Mon, Mar 30, 2026 at 10:57:44AM +0200, Wolfram Sang wrote:
> 
> > > Sure thing. Is there a dedicated mailing-list or better email address I
> > > can add?
> > 
> > Not yet, but I think of creating one.
> 
> Until that exists, shall I use your email to add Reported-by tags? In
> another of my series, Sashiko found valid issues which already existed
> before my series. A tag would be proper, I'd think?

I was thinking about proposing some tagging convention such as:

   Suggested-by: Sashiko:Gemini 3.1 Pro
or
   Reviewed-by: Sashiko:Gemini 3.1 Pro

to Documentation/process/coding-assistants.rst.  Alas, neither is
perfect.

Suggested-by: is generlly used when someone inspires a particular
commit.  This might apply if Sashiko found a problem as an incidental
finding, which we then fixed in a subsequent commit.  An example of
this might be[1], or in the case which you suggested above.  But what
if we just fixed one of the issues raised by Sashiko in an earlier
version of the commit.  In that case, Suggested-by: doesn't seem to be
a perfect fit.

[1] https://lore.kernel.org/r/20260327063330.1312426-1-tytso@mit.edu/

Reviewed-by: is generally only applicable once *all* of the issues
identified by the reviewer has been resolved, and it's not clear this
is applicable if not all of the issues rasied by Sashiko were
resolved.  In some cases, these might be false positives, but in the
case of a human reviewer, the human reviewer agrees before saying,
"You can add Reviewed-by: ..." to the commit.  Unfortunately, it will
probably be a while before LLM's have that kind of agency / consciousness.  :-)

What do folks think?  How should we codify a way of giving Sashiko
credit for issues that it has raised?  (Assuming we should, but
hopefully that's not controversial.)

          	     		       	      - Ted