Re: [PATCH v6 0/4] ppc: Enable full Xen build

[George Dunlap <george.dunlap@xxxxxxxxx>]

On Mon, Sep 18, 2023 at 6:27 PM Shawn Anastasio
<sanastasio@xxxxxxxxxxxxxxxxxxxxx> wrote:
>
> On 9/18/23 8:19 AM, Jan Beulich wrote:
> > On 14.09.2023 21:03, Shawn Anastasio wrote:
> >> Shawn Anastasio (4):
> >>   xen/ppc: Implement bitops.h
> >>   xen/ppc: Define minimal stub headers required for full build
> >
> > Compilation fails after applying this.
> >
> >>   xen/ppc: Add stub function and symbol definitions
> >
> > Continuing nevertheless, linking fails after this.
> >
> >>   xen/ppc: Enable full Xen build
> >
> > Things build okay for me when the full series is applied. Generally we
> > wouldn't deliberately break the build between any two patches; doing so
> > may be okay here (except I guest CI's build-each-commit would be upset),
> > but I'll do so only upon explicit request (and with no-one else objecting).
> >
>
> Sorry about that. Going forward I'll take more care to ensure that
> partially-applied series still build correctly. For this series though,
> if you could make an exception it would be appreciated.

What would be the reason for the exception?

We don't want to follow the rules just for the rules' sake, but the
rule is there for a reason: primarily to keep bisection working.  Not
sure of osstest is testing the PPC build yet, but if it were, then
this sort of thing would make it more difficult for the automatic
bisector to find regressions in other parts of the code.  Having
non-building patches can also confuse "archaeologists" -- people a few
years hence who are trying to understand what the code does.

Is there a reason that this series would be particularly difficult to
reorganize in a way that would keep it building?  (Haven't looked at
it in detail.)

 -George