Re: linux-next: build failure after merge of the final tree

[Thomas Gleixner <tglx@linutronix.de>]

On Fri, 20 May 2011, Linus Torvalds wrote:
> On Thu, May 19, 2011 at 11:12 PM, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> >
> > After merging the final tree, today's linux-next build (sparc32 defconfig)
> > failed like this:
> 
> Hmm. So I had actually done a "allyesconfig" build on x86, which
> annoys me. Because it means that the extra "let's compile everything
> to make sure I didn't break anything" was just almost totally
> worthless.
> 
> What seems to be happening is that the x86 <asm/uaccess.h> include
> ends up getting the <linux/prefetch.h>.
> 
> I have *no* idea why x86 does that, but x86 wants prefetch.h *so* much
> that it actually includes it first in <asm/uaccess.h> and then *again*
> in each of the 32/64-bit specific <asm/uaccess_[32,64].h> header
> files.
> 
> That seems a bit excessive. I don't think x86 should include
> <linux/prefetch.h> at all, since (a) it doesn't actually use any of
> it, and (b) it ended up hiding this problem from me.
> 
> Thomas, Ingo, Peter: would you be willing to just remove that stupid
> header file inclusion and fix up the fallout? Instead of having these
> one-by-one patches that come from Stephen testing out breakage on
> other architectures that x86 simply hid due to its odd include files?

Removed it, but it does not break anything on x86 because
linux/thread_info.h includes asm/thread_info.h which includes
asm/processor.h on x86 for non obvious reasons.

Thanks,

	tglx