Re: [PATCH 18/22] Use allocation flags as an index to the zonewatermark

[Mel Gorman]

On Thu, Apr 23, 2009 at 09:27:15AM +0900, KOSAKI Motohiro wrote:
> > On Wed, 2009-04-22 at 18:14 +0100, Mel Gorman wrote:
> > > Preference of taste really. When I started a conversion to accessors, it
> > > changed something recognised to something new that looked uglier to me.
> > > Only one place cares about the union enough to access is via an array so
> > > why spread it everywhere.
> > 
> > Personally, I'd say for consistency.  Someone looking at both forms
> > wouldn't necessarily know that they refer to the same variables unless
> > they know about the union.
> 
> for just clalification...
> 
> AFAIK, C language specification don't gurantee point same value.
> compiler can insert pad between struct-member and member, but not insert
> into array.
> 

Considering that they are the same type for elements and arrays, I
didn't think padding would ever be a problem.

> However, all gcc version don't do that. I think. but perhaps I missed
> some minor gcc release..
> 
> So, I also like Dave's idea. but it only personal feeling.
> 

The tide is against me on this one :).

How about I roll a patch on top of this set that replaces the union by
calling all sites? I figure that patch will go through a few revisions before
people are happy with the final API. However, as the patch wouldn't change
functionality, I'd like to see this series getting wider testing if possible. The
replace-union-with-single-array patch can be easily folded in then when
it settles.

Sound like a plan?

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/