Re: [PATCH 02/11] atyfb_base compile fix for CONFIG_PCI=n

[Andrew Morton <akpm@osdl.org>]

Greg KH <greg@kroah.com> wrote:
>
> On Wed, May 24, 2006 at 06:33:27PM -0700, Andrew Morton wrote:
> > zippel@linux-m68k.org wrote:
> > >
> > >  The atyfb_driver structure is only available if CONFIG_PCI is set.
> > > 
> > >  Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
> > > 
> > >  ---
> > > 
> > >   drivers/video/aty/atyfb_base.c |    4 ++++
> > >   1 file changed, 4 insertions(+)
> > > 
> > >  Index: linux-2.6-mm/drivers/video/aty/atyfb_base.c
> > >  ===================================================================
> > >  --- linux-2.6-mm.orig/drivers/video/aty/atyfb_base.c
> > >  +++ linux-2.6-mm/drivers/video/aty/atyfb_base.c
> > >  @@ -3861,7 +3861,9 @@ static int __init atyfb_init(void)
> > >       atyfb_setup(option);
> > >   #endif
> > >   
> > >  +#ifdef CONFIG_PCI
> > >       pci_register_driver(&atyfb_driver);
> > >  +#endif
> > >   #ifdef CONFIG_ATARI
> > >       atyfb_atari_probe();
> > >   #endif
> > >  @@ -3870,7 +3872,9 @@ static int __init atyfb_init(void)
> > >   
> > >   static void __exit atyfb_exit(void)
> > >   {
> > >  +#ifdef CONFIG_PCI
> > >   	pci_unregister_driver(&atyfb_driver);
> > >  +#endif
> > >   }
> > 
> > bah.  If pci_register_driver() was a macro we wouldn't need to do this all
> > over the place.
> 
> Yes, this can be fixed easily in the pci.h header file, all other pci
> functions are stubbed out properly if CONFIG_PCI is not enabled.  These
> should be too.
> 

Well pci_register_driver() and pci_unregister_driver() do have stubs.  But
they're static-inlines, hence they reference their argument, hence the
above ifdefs.

But if the pci_register_driver() and pci_unregister_driver() stubs were
macros which do not reference their argument, the above ifdefs aren't
needed.