Re: HEADS UP: MIT KRB5 1.22.1 Import

From: Cy Schubert <Cy.Schubert_at_cschubert.com>
Date: Thu, 21 Aug 2025 16:40:30 UTC
In message <20250821162750.D9BCC307@slippy.cwsent.com>, Cy Schubert writes:
> In message <CANCZdfo9_D-D43JnoTxt1ivN_SZ2smpHQTJRL2oqp9q+YZEqfQ@mail.gmail.c
> om>
> , Warner Losh writes:
> > --0000000000004b0148063ce187a5
> > Content-Type: text/plain; charset="UTF-8"
> > Content-Transfer-Encoding: quoted-printable
> >
> > On Thu, Aug 21, 2025 at 9:10=E2=80=AFAM Cy Schubert <Cy.Schubert@cschubert.
> =
> > com>
> > wrote:
> >
> > > In message <aKcqsXIyYobdRIlY@nuc>, Mark Johnston writes:
> > > > On Thu, Aug 21, 2025 at 02:36:15AM -0700, Cy Schubert wrote:
> > > > > In message <86frdlkw30.fsf@ltc.des.dev>,
> > > =3D?utf-8?Q?Dag-Erling_Sm=3DC3=3DB8rgrav
> > > > ?
> > > > > =3D w
> > > > > rites:
> > > > > > Cy Schubert <Cy.Schubert@cschubert.com> writes:
> > > > > > > KRB5 1.22.0 released with a GSSAPI bug (CVE-2025-57736). The poin
> =
> > t
> > > patc
> > > > h=3D
> > > > > > =3D20
> > > > > > > was applied on Aug 19 (f96110babbe1). 1.22.1 contains the point
> > > patch p
> > > > lu=3D
> > > > > > s=3D20
> > > > > > > an additional GSSAPI patch. I intend to commit it to HEAD after a
> > > few d
> > > > ay=3D
> > > > > > s=3D20
> > > > > > > of testing on Aug 25.
> > > > > >
> > > > > > Please post your patch for review (using git arc, as described in t
> =
> > he
> > > > > > committers guide) so others can test it too.
> > > > >
> > > > > git arc does not work with vendor merges. There is a way to trick it
> > > though
> > > > .
> > > >
> > > > What problem do you run into exactly?  Could you provide a recipe for m
> =
> > e
> > > > to reproduce it locally?
> > >
> > > It's a vendor import. It's the same problem we have with git rebase and
> > > git
> > > diff of a vendor import. (This is why vendor imports use a special
> > > procedure [see committers guide].)
> > >
> >
> > I think you just need to tell arc to use the first parent....
>
> That would be the commit hash in the vendor branch?
>
> commit dbbb504963792cabf7a2ef862dbb92b254f56782 (HEAD -> krb5-1221)
> Merge: 458ac2b6e445 e5fe63eaf1d3
> Author:     Cy Schubert <cy@FreeBSD.org>
> AuthorDate: Wed Aug 20 22:38:21 2025 -0700
> Commit:     Cy Schubert <cy@FreeBSD.org>
> CommitDate: Wed Aug 20 22:40:52 2025 -0700
>
>     krb5: Update to 1.22.1
>     
>     Merge commit 'e5fe63eaf1d35ebbeac17eeed04cf873fbb9b3da' into main
>
> commit e5fe63eaf1d35ebbeac17eeed04cf873fbb9b3da (tag: vendor/krb5/1.22.1, 
> freebsd/vendor/krb5, vendor/krb5)
> Author:     Cy Schubert <cy@FreeBSD.org>
> AuthorDate: Wed Aug 20 22:22:31 2025 -0700
> Commit:     Cy Schubert <cy@FreeBSD.org>
> CommitDate: Wed Aug 20 22:22:47 2025 -0700
>
>     krb5: Import MIT 1.22.1
>     
>     Security:       CVE-2025-57736
>
>
> Note that a git format-patch -2 doesn't include the 
> e5fe63eaf1d35ebbeac17eeed04cf873fbb9b3da commit.
>
> I don't understand why the vendor commit no longer appears in history after 
> the vendor branch merge has been pushed.

I used the vendor branch commit hash. As with git-diff the files are rooted 
from contrib/krb5 not from the root of the source tree. This must be a 
subtree merge thing.


-- 
Cheers,
Cy Schubert <Cy.Schubert@cschubert.com>
FreeBSD UNIX:  <cy@FreeBSD.org>   Web:  https://FreeBSD.org
NTP:           <cy@nwtime.org>    Web:  https://nwtime.org

			e**(i*pi)+1=0