[RFC 0/6] Regressions for "imply" behavior change

[Jason Gunthorpe]

On Thu, Apr 16, 2020 at 05:58:31PM +0200, Arnd Bergmann wrote:
> On Thu, Apr 16, 2020 at 4:52 PM Jason Gunthorpe <jgg at ziepe.ca> wrote:
> > On Thu, Apr 16, 2020 at 02:38:50PM +0200, Arnd Bergmann wrote:
> > > On Thu, Apr 16, 2020 at 12:17 PM Jani Nikula <jani.nikula at linux.intel.com> wrote:
> > > > Of course, this is all just talk until someone(tm) posts a patch
> > > > actually making the change. I've looked at the kconfig tool sources
> > > > before; not going to make the same mistake again.
> > >
> > > Right. OTOH whoever implements it gets to pick the color of the
> > > bikeshed. ;-)
> >
> > I hope someone takes it up, especially now that imply, which
> > apparently used to do this, doesn't any more :)
> 
> The old 'imply' was something completely different, it was more of a
> 'try to select if you can so we can assume it's there, but give up
> if it can only be a module and we need it to be built-in".

But it seems to have done this as a side-effect, and drivers were
relying on that, otherwise this series wouldn't exist..

Jason