Re: [GIT PULL] x86: use arch/x86/include

[Ingo Molnar <mingo@elte.hu>]

* Sam Ravnborg <sam@ravnborg.org> wrote:

> > hm, we have a _ton_ of changes to include files queued up already, 
> > so this is rather inconvenient.
>
> git should cope with this if changes come in via git. I do not know 
> about "git am" applied patches.

the pending commits are ordinary commits and there's little difference 
between a leaf commit that came via git and one that came via emails.

(the difference between Git and email is for more complex ops like 
renames, merges - i.e. more abstract and multi-commit operations.)

But such more complex scenarios are not what i'm talking about. We've 
got this many leaf commits in include/asm-x86/ at the moment:

   308 files changed, 3025 insertions(+), 2025 deletions(-)

... and pulling your rename generates almost 50 conflicts. (Also, i've 
got to hunt down all scripts that somehow rely on the location on 
include/asm-x86.)

> > I missed the discussion on this, what's the point of renaming all 
> > these files?
> 
> It has been discussed many times to keep arch and arch include files
> under arch/.
> Lately Linus outlined this:
> 
>     http://kerneltrap.org/mailarchive/linux-kernel/2008/5/21/1903924
> 
> I took the ball and did the kbuild side of this so we could do this 
> gradually.
>
> sparc are already converted. sh has it ready to be pulled and others
> are playing with it.
> 
> How we do it in the best way for such a fragile codebase as x86 is I 
> am not sure. [...]

huh, fragile codebase? Is that a flamebait? :-) What do you mean 
exactly?

The timing problems come from the fact that 90% of Linux development and 
95% of Linux testing happens on x86. So we've already got a ton of stuff 
queued up for the next merge window. (and some cleanups for this cycle 
as well)

But fortunately you've scripted all this, so i guess we can do it at any 
stage. Could you send the script we should run?

	Ingo