On Fri, 2026-11-06 at 15:01 +0100, Helmut Grohne wrote:
> Hi,
> 
> Bastian disagrees with me that there is a bug here off list.
> 
> On Fri, Nov 06, 2026 at 12:05:26PM +0100, Helmut Grohne wrote:
> > linux-image-6.12.115+deb13.1-riscv64-dbg has an undeclared file
> > conflict.  This may result in an unpack error from dpkg.
> > 
> > The file
> > /usr/lib/debug/.build-id/90/772e7b336781d21fcb7f165943e58430842d86.debug
> > is contained in the packages
> >  * linux-image-6.12.115+deb13-riscv64-dbg/6.12.115-1 as present in 
> > trixie-security
> >  * linux-image-6.12.115+deb13.1-riscv64-dbg/6.12.115-2 as present in 
> > trixie-security
> > 
> > These packages can be unpacked concurrently, because there is no
> > relevant Replaces or Conflicts relation.  Attempting to unpack these
> > packages concurrently results in an unpack error from dpkg, because none
> > of the packages installs a diversion for the affected file.
> 
> The practical instance of this bug may be unsolvable now. We're not
> uploading a new package to trixie-security to add Replaces or similar.
> However, there is a fundamental problem here that resurfaces now and
> then: Colliding build ids. This is not exactly a new problem. It is even
> documented at
> https://docs.kernel.org/kbuild/reproducible-builds.html#debug-info-conflicts.

By me.

> So to me it looks like our CONFIG_BUILD_SALT is not unique enough and
> this is the problem actually being reported here.

I believe we do generate a unique CONFIG_BUILD_SALT for each package
version and flavour.  The problem is that each architecture has to
explicitly include that string in its vDSO, and some of them weren't
doing that.

Bastian fixed this upstream in Linux 7.2.  Which is why the report was
closed.

Ben.

-- 
Ben Hutchings
For every action, there is an equal and opposite criticism. - Harrison

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to