[SRU F/G/U 0/3] Enforce all annotations

Thadeu Lima de Souza Cascardo cascardo at canonical.com
Wed May 20 10:23:12 UTC 2020


On Wed, May 20, 2020 at 09:12:05AM +0200, Stefan Bader wrote:
> On 18.05.20 15:36, Thadeu Lima de Souza Cascardo wrote:
> > BugLink: https://bugs.launchpad.net/bugs/1879327
> > 
> > Start enforcing all config annotations. This is implemented as an opt-in
> > for now, that each branch may set as default. After all branches are fixed
> > up to set it by default, we can turn it into an opt-out.
> > 
> > It's set to true for the master branch as it builds fine with Focal.
> > Unstable, however, might break, but I am sending the opt-in patch for
> > Unstable as well, nonetheless.
> > 
> > Thadeu Lima de Souza Cascardo (3):
> >   UBUNTU: [Config]: do not enforce CONFIG_VERSION_SIGNATURE
> >   UBUNTU: [Config]: prepare to enforce all
> >   UBUNTU: [Config]: enforce all config options
> > 
> >  debian.master/rules.d/hooks.mk   |  1 +
> >  debian/rules                     |  1 +
> >  debian/rules.d/4-checks.mk       |  3 ++-
> >  debian/scripts/config-check      | 12 ++++++++----
> >  debian/scripts/misc/kernelconfig |  2 +-
> >  5 files changed, 13 insertions(+), 6 deletions(-)
> >  create mode 100644 debian.master/rules.d/hooks.mk
> > 
> IMO this should be first getting applied to G/U and when there are no surprises,
> this could be added to Focal as a next step.
> 
> -Stefan
> 
> 

One of the challenges here is that unstable development process is more prone
to failures, as it's constantly rebased on top of new upstream versions, full
of config changes.

This is really a stable feature, preventing config regressions. I could submit
it to all series, but thought restricting it to focal was a better idea, as it
would: 1) restrict to only one series; 2) focal master already has all config
and annotations match.

Of course, this should be applied first to unstable, but not sure that waiting for
"no surprises" there will work.

This is certainly going to change some of the stable process, potentially
causing more build failures than before, as config options change without the
respective annotations change. Isn't that what we want it to do? In order to
catch unreviewed changes or even regressions?

Cascardo.



More information about the kernel-team mailing list