Re: [PATCH b4 0/2] Detect naked cover letters that lack a 0/N prefix and diffstat

On 2026-07-27 16:37:45+02:00, Christian Brauner wrote:
> > LoreMailbox only recognizes a cover letter when the message carries 
> > an
> > explicit zero counter ([PATCH 0/N]) or, via the in-reply-to fallback in
> > get_series(), when the thread root carries a diffstat.  A "naked" cover --
> > one with neither a [PATCH 0/N] subject prefix nor a diffstat in the body --
> > matches neither test, so it lands in self.unknowns and is dropped, and
> > LoreSeries.has_cover stays False.
> > 
> > The visible fallout is in shazam -M, which then reports "No cover letter
> > provided by the author" and links the first patch instead of the real
> > cover.  This happens whenever an author writes a short prose cover with no
> > diffstat, e.g. Christoph Hellwig's "don't build bios/contexts over multiple
> > iomaps" series:
> > 
> >   https://lore.kernel.org/linux-xfs/[email protected]
> > 
> > Patch 1 extends the in-reply-to fallback to also accept the parent as the
> > cover when it is a non-patch, same-author thread root (in_reply_to is None
> > and not itself a reply).  Requiring the same author and the thread root
> > avoids mistaking an unrelated message the series was merely sent
> > in-reply-to -- a third party's bug report, say -- for a cover letter.
> > 
> > Patch 2 adds positive (naked single-patch cover) and negative (patch sent
> > in-reply-to a stranger's bug report) fixtures and a test.
> > 
> > Signed-off-by: Christian Brauner (Amutable) <[email protected]>
> > ---
> 
> A solution to this problem would also be nice as Christoph keeps sending
> me patch series with naked cover letters and I have to fight with b4 to
> apply them correctly. :)

I will take this, but I will grumpily point out that it's the second 
time today that I'm adding special handling for patch series generated 
basically the same way as in 1999.

-K