Re: [RFC] [PATCH] virtual memmap on sparsemem v3 [3/4] static virtual mem_map

[Andrew Morton <akpm@osdl.org>]

On Sat, 9 Dec 2006 11:49:50 +0900
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote:

> On Fri, 8 Dec 2006 16:30:20 -0800
> Andrew Morton <akpm@osdl.org> wrote:
> 
> > > +#ifdef CONFIG_SPARSEMEM_VMEMMAP_STATIC
> > > +#include <linux/mm_types.h>
> > > +extern struct page mem_map[];
> > > +#else
> > >  extern struct page* mem_map;
> > >  #endif
> > > +#endif
> > 
> > This looks rather unpleasant - what went wrong here?
> > 
> definition of 'struct page' is necessary before declearing array[]. (for gcc-4.0)

hm, OK, that declaration needs all of `struct page' in scope.

> 
> > Would prefer to unconditionally include the header file - conditional inclusions
> > like this can cause compile failures when someone changes a config option.  They
> > generally raise the complexity level.
> >
> Okay.
> Now, forward declearation of 'struct page' is in mmzone.h. 
> I'll remove it and include mm_types.h instead of it.
> If someone says "Don't do that", I'll look for anothere way.
> 

This header needs mm_types.h, so including it is certainly OK - there's no
choice.  But I think it'd be better to include mm_types.h outside of any
ifdefs.  Just stick the #include at the start of the file as usual.