Re: [RFC] [PATCH -mm] ASIC3 driver

[Thomas Gleixner <tglx@linutronix.de>]

On Fri, 19 Oct 2007, Andrew Morton wrote:

> On Fri, 19 Oct 2007 12:53:00 +0200 Samuel Ortiz <sameo@openedhand.com> wrote:
> 
> > On Thu, Oct 18, 2007 at 03:05:44PM -0700, Andrew Morton wrote:
> > > On Thu, 18 Oct 2007 11:12:41 +0200
> > > Samuel Ortiz <sameo@openedhand.com> wrote:
> > 
> > > > +#include <linux/module.h>
> > > > +#include <linux/version.h>
> > > > +#include <linux/irq.h>
> > > 
> > > Please see the large comment at the top of linux/irq.h.  I believe this
> > > driver will fial to compile on at least arm.
> > It doesn't build as a module, since we need the irq.h symbols.
> > I changed MFD_ASIC3 to bool. I somehow feel that this is not the cleanest
> > solution, but OTOH I think that dynamically adding IRQs and GPIOs to an
> > embedded board doesn't make much sense.
> 
> We seem to have miscommunicated here.  <linux/irq.h> contains references to
> things which only some architectures actually implement.  I don't know
> which architectures those are, but it includes common ones like x86, so
> it's a real trap.  I recall it does not include arm, so your code might
> break on arm.
> 
> At least, that's what's _supposed_ to happen: I just compiled and linked
> this driver into an ARM kernel with no problems so now I'm all confused as
> to what the problem was.
> 
> Oh well, we'll see...

We obviously never removed the big fat warning, which was valid before
the ARM to generic irq conversion.

    tglx