On Tuesday 21 July 2009, Manuel Lauss wrote:
> >> -#ifdef CONFIG_PM
> > Won't the removal of this test cause a build failure if CONFIG_PM is
> > not set? If the removal of the test is safe, this should IMHO at
> > least be explained in the commit message.
> No, it builds just fine without CONFIG_PM; it was there to shave off a
> few bytes from the kernel image. But not everyone tests this driver
> with CONFIG_PM=y, because apparently noone really needed PM on
> this platform (Alchemy), and a full build of most of the boards using
> this driver fails with PM enabled.
> This way the PM methods at least get a compile-test in the non-pm case.
Not sure that is a sufficiently valid argument. In any case it *is* a
separate change to "dev_pm_ops conversion" so it really should at least
be documented and justified in the commit log.
> I like what Magnus Damm did for some of the SuperH drivers:
> #ifdef CONFIG_PM
> #define DRIVER_PM_OPS (&driver_pm_ops)
> #define DRIVER_PM_OPS NULL
Yes, that's quite elegant.
> I'd like to keep the pm stuff enabled at all times since it doesn't
> hurt in the non-pm case and if kernel size becomes a problem I can add
> the #defines back.
I guess that's up to the maintainers of the mips port.