don't base trees on 5.12-rc1

[Daniel Vetter]

On Wed, Mar 10, 2021 at 8:11 PM Dave Airlie <airlied at gmail.com> wrote:
>
> On Wed, 10 Mar 2021 at 17:48, Maxime Ripard <maxime at cerno.tech> wrote:
> >
> > Hi Dave,
> >
> > On Wed, Mar 10, 2021 at 09:50:29AM +1000, Dave Airlie wrote:
> > > I'm mostly sending this to the -misc maintainers because
> > > drm-misc-fixes is based on rc1 at present.
> > >
> > > This needs to be *rebased* not merged up to 5.12-rc2. Merging will
> > > still have the bad landmine commits in the bisect history. This is a
> > > very special case.
> >
> > I'm sorry, I'm not entirely sure I get this. -rc1 is still in the -rc2
> > history, so how would that change anything in the bisect history?
> >
>
> We can't get rid of the bad commit range, we can reduce the amount of
> times someone accidentally bisects into it, by not using it as a base
> commit for future changes.
>
> If in the future a bisect happens to want to test one of the patches
> in drm-misc-fixes that is based on rc1, it will land the user with an
> rc1 test kernel and could eat their swapfile/disk. We can avoid that
> problem by not using rc1 as a base for drm-misc-fixes.
>
> We can't avoid them bisecting into the broken commits between when
> this landed and was fixed, but rebasing trees can minimise the chances
> of this when bisecting other changesets.

Same for backmerge, backmerge -rc2, not -rc1. I think there's a
request for backmerge pending from Noralf anyway. Also for any topic
branch and all that.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch