Re: CONFIG_OPTIMIZE_INLINING

[Sam Ravnborg <sam@ravnborg.org>]

On Sun, Apr 27, 2008 at 02:26:00AM -0400, Christoph Hellwig wrote:
> On Sun, Apr 27, 2008 at 07:59:43AM +0200, Ingo Molnar wrote:
> > here, two months ago:
> > 
> >   http://lkml.org/lkml/2008/3/3/122
> > 
> > i made it x86-only due to this FUD that went against it:
> > 
> >   "messing with a global #define in a way that the results on 24 
> >    architectures with 7 different releases of gcc would be 
> >    unpredictable."
> > 
> > ... as i saw no reason why this feature, which i found rather useful, 
> > should be delayed another year or so. I'd be more than happy to promote 
> > this feature back to lib/Kconfig.debug, sparc64 interest would make that 
> > a strong argument.
> > 
> > and here is the pull request that i posted to lkml, with the full patch 
> > included as well:
> > 
> >   http://www.gossamer-threads.com/lists/linux/kernel/911104?page=last
> 
> This still does not explain why it's two commits.   It also doesn't
> explain the horrible cpp and ifdef abuse in there - having a config
> symbol in one architecture only and then a cpp symbol to prevent the ill
> effects of it in one signle architecture is simply utterly braindead,
> sorry.
> 
> This kind of thing should be discussed at least on linux-arch and input
> from at least partially wise people would have led too a much better
> solution.  Like, umm moving the option to a global config file and let
> the architecture default it to y/n where needed.

I just posted a patch to linux arch that to some extent do as you sketch above.
Then I hope some of the "partially wise" people will nack or ack or suggest
better ways to do it.

	Sam